FACULTY OF INFORMATION STUDIES IN NOVO MESTO D O C T O R A L D I S S E R T A T I O N JERNEJ AGREŽ

Transkripcija

1 FACULTY OF INFORMATION STUDIES IN NOVO MESTO D O C T O R A L D I S S E R T A T I O N JERNEJ AGREŽ

2

3 FACULTY OF INFORMATION STUDIES IN NOVO MESTO DOCTORAL DISSERTATION FUTURE KNOWLEDGE AND PROCESS PATTERN RECOGNITION IN WEAKLY DEFINED ORGANIZATIONAL FORMATIONS Adviser: Prof. dr. Nadja Damij PhD Novo mesto, May 2016 Jernej Agrež

4 IZJAVA O AVTORSTVU Podpisani Jernej Agrež, študent FIŠ Novo mesto, v skladu z določili statuta FIŠ izjavljam: da sem doktorsko disertacijo pripravljal samostojno na podlagi virov, ki so navedeni v doktorski disertaciji, da dovoljujem objavo doktorske disertacije v polnem tekstu, v prostem dostopu, na spletni strani FIŠ oz. v digitalni knjižnici FIŠ, da je doktorska disertacija, ki sem jo oddal v elektronski obliki identična tiskani verziji, da je doktorska disertacija lektorirana. V Novem mestu, dne Podpis avtorja

5 When one jumps over the edge, one is bound to land somewhere. D. H. Lawrence Hvala mentorici, profesorici Nadji Damij, in ekipi FIŠ, ki je poskrbela, da ta skok ne bo nikoli pozabljen, kolegom reševalcem in ostalim kolegom iz civilne zaščite, ki so bili neizogibno vpleteni ob nastajanju tega dela, ter ne nazadnje, hvala moji domači posadki, ki je skočila in pristala skupaj z mano.

6

7 ABSTRACT Customer knowledge as an influence on service provider s business processes during emergencies is an interesting and relevant research topic. The aim of this research was to identify possible emergency response process improvements based on existing learning processes, and possible future knowledge exchange improvements. I reached the predefined objective, i.e. to identify how customer knowledge influences providers business processes and consequently triggers social impact, by conducting extensive work on all five parts of the thesis. I confirmed that the customer-provider relation based on customer knowledge exists and appears in several different forms. Finally, I identified possible improvements of the research and possibilities for further research by applying the designed methodological framework network analysis on a broader geographical scale. A significant contribution of this thesis is the integration of three fields of management: business process management, knowledge management, and disaster management. The research focuses on identifying knowledge-based customer-provider relations by emphasizing customer knowledge as an influence on provider s business processes and possible consequential social impact. By studying these relations, I went beyond the usual comfort zone of conventional formal organizations and employed a loosely coupled system which emerges as a flood response mechanism during flood events. KEY WORDS: business process management, knowledge management, disaster management, public safety, loosely coupled systems, customer knowledge, flood events.

8

9 POVZETEK Znanje potrošnikov, ki nastopi kot vpliv na poslovne procese ponudnika storitev, v stanju izrednih razmer, je zanimivo in relevantno raziskovalno področje. Namen raziskave je bil identificirati možne izboljšave procesa odziva, ob naravnih nesrečah, ki bi izhajale iz obstoječih učnih procesov in morebitnih bodočih izboljšavah izmenjave znanja. Zadan cilj, ugotoviti kako znanje potrošnikov vpliva na poslovne procese ponudnika storitev in kako se le-to odrazi v družbenem okolju, sem dosegel s poglobljeno raziskavo na petih segmentih tega dela. Potrdil sem, da se odnos med potrošnikom in ponudnikom, ki izhaja iz znanja potrošnika, pojavlja v različnih oblikah. Nenazadnje sem nanizal tudi možnosti za izboljšavo raziskave in možnosti za nadalnje raziskovalno delo, preko uporabe metod za analizo kompleksnih omrežij in izvedbo raziskave na širšem geografskem območju. Pomemben znanstveni doprinos tega dela je združitev upravljanja poslovnih procesov, upravljanja znanja in obvladovanja naravnih nesreč. Raziskava je usmerjena na indentifikacijo odnosa med potrošnikom in ponudnikom, ki temelji na znanju potrošnika kot nosilcu vpliva na poslovne procese ponudnika, takšen vpliva pa se lahko dotakne tudi socialnega okolja. Z raziskovanjem tovrstnih odnosov, sem presegel okvir formalne organizacije in svoje delo apliciral na ohlapen organizacijski sistem, ki se vzpostavi kot zaščitni mehanizem ob nastanku poplav. KLJUČNE BESEDE: upravljanje poslovnih procesov, upravljanje znanja, obvladovanje naravnih nesreč, javna varnost, ohlapni sistemi, znanje potrošnikov, poplave.

10

11 CONTENT 1 INTRODUCTION Research problem Research problem background Relevance of the research problem Motivation Hypotheses Methodology Research objectives Scientific contribution Structure of the thesis KEY CONCEPTS AND LITERATURE REVIEW System and organization Loose coupling Looseness and sustainability, looseness and continuity Business process management Business process architecture Business process improvement Business process modelling Business process simulations Knowledge management Organizational knowledge Tacit, explicit and experiential knowledge in the loosely coupled system Experiential community learning Loosely coupled process architecture and organizational learning Open knowledge management approach to a loosely coupled system... 51

12

13 2.3.6 Knowledge mapping EMPIRICAL RESEARCH From theory to practical applications H1 H4, Background information H1 H4, Data Testing Hypotesis Methodology Results Testing Hypothesis Methodology Results Testing Hypothesis Methodology Results Testing Hypothesis Methodology Results Testing Hypothesis Background information Data Methodology Results CONCLUSION Research problem Hypotheses Hypothesis Hypothesis

14

15 4.2.3 Hypothesis Hypothesis Hypothesis Methodology Research objectives Scientific contribution Practical application and research limitations Possible improvements and further research REFERENCES SUBJECT INDEX INDEX OF AUTHORS BODOČE ZNANJE IN PREPOZNAVA PROCESNIH VZORCEV ZNOTRAJ NEJASNO DEFINIRANIH ORGANIZACIJSKIH STRUKTUR (obsežen povzetek) APPENDICES

16

17 LIST OF TABLES Table 1.1: Sample presentation Table 2.1: Differences between loosely coupled and tight systems Table 2.2: Comparison of the subsystems Table 3.1: Entity data Table 3.2: Entity data Table 3.3: Flood response activities Table 3.4: Hydrological data Table 3.5: Meteorological data Table 3.6: Adjusted TAD activity table Table 3.7: Adjusted TAD activity table Table 3.8: Classification confusion matrix Table 3.9: Detailed accuracy by class Table 3.10: Semi-structured interview framework Table 3.11: Interview data nominal scale Table 3.12: Quantitative results of the semi-structured interview Table 3.13: Quantitative results of the semi-structured interview Table 3.14: Quantitative results of the semi-structured interview Table 3.15: Mean, standard deviation values and paired t-test Table 3.16: River flow data, multivariate analysis of variance Table 3.17: Rainfall data, multivariate analysis of variance Table 3.18: Weighting criteria Table 3.19: Mean values, standard deviation values and paired t-test of AS-IS and TO-BE process states Table 3.20: Transaction count Table 3.21: Transaction count, mean values and standard deviation Table 3.22: Importance of components Table 3.23: Attributes placed in the framework of principal components Table 3.24: Chi-Square test Table 3.25: Identified entities Table 3.26: TAD extension, knowledge table

18

19 Table 3.27: Knowledge simulation, transaction count Table 3.28: Extracted important knowledge and key knowledge sources Table 3.29: Multivariate tests Table 3.30: Tests of between-subjects effects

20

21 LIST OF FIGURES Figure 2.1: Five groups of posteriori knowledge Figure 3.1: The Lower Sava statistical region, Slovenia Figure 3.2: Geographic region: The Lower Sava Valley, Slovenia Figure 3.3: Flood affected communities during the 2012 floods (Dolenjski list, 2012) Figure 3.4: Structure of the simulation Figure 3.5: Pareto chart comparison of the original data and simulation output Figure 3.6: Response process workload of the primary response force Figure 3.7: Response process workload of the secondary response force Figure 3.8: Response process workload of the tertiary response force Figure 3.9: Designed learning evaluation fuzzy system Figure 3.10: Flow rate measurements during flood events Figure 3.11: Rainfall measurements during flood events Figure 3.12: Fuzzy system plot of the most successful learning communities Figure 3.13: Correlations of attribute values, typical for entities from successful learning communities Figure 3.14: Process optimization algorithm design Figure 3.15: Scatter matrix of AS-IS and TO-BE process states data Figure 3.16: Flood event response process model Figure 3.17: Comparison between proportions of mean values and standard deviation values Figure 3.18: Comparison of knowledge exchange transactions Figure 3.19: Multiple correspondence analysis plot Figure 3.20: Principal component analysis plot Figure 3.21: TAD Activity table extended with knowledge mapping notation Figure 3.22: Numerical distribution of process transactions Figure 3.23: Numerical distribution of process transaction correlations between different simulation scenarios Figure 3.24: Model of significant knowledge, and knowledge sources in the process

22

23 1 INTRODUCTION Business processes happen all around us: in a grocery store, in a kindergarten, at a construction site, in sport clubs, in multinational corporations, etc. However, many people involved in business processes do not recognize them as such. By focusing only on the sets of activities that are easier to recognize, more tangible and simpler to approach, it is difficult to see the wide perspective of an operation or an organization and to identify the business processes linked to it. Similar phenomena could be attributed to knowledge, which is mostly defined as something one learns at school: mathematics, geography, languages, biology, etc. By understanding knowledge in the same way as the afore-mentioned activities very specific, focused and explicit we miss the wide aspect of knowledge as an influence which could bring about and support change in communities, an influence surrounding both business processes and accumulated knowledge. A deliberation of how knowledge influences business processes easily yields several ideas such as: better trained workers will be able to conduct their work more efficiently, consequently leading to greater success of the business process; a better educated CEO will have the capacity to manage the company more wisely and prudently, consequently improving the company s business processes. But when trying to determine how customer knowledge influences business processes, one of the first things that come to mind is the reverse information channel about customer satisfaction and experience with the product or service. The aim of the thesis is to establish how customer knowledge influences business processes and how this affects the surrounding communities. Moreover, I want to project the future knowledge into the simulated model of the assessed case and determine what changes arise from the knowledge that has not yet occurred. To this end, I decided to apply the research work on a loosely coupled organizational system which differs from conventional formal organization in several aspects. Among several examples of a loosely coupled system (informal NGO networks, project consortiums, multidisciplinary business clusters, etc.), I decided to assess a flood response system as an interesting representative of a loosely coupled system. For verification, I chose a network of individuals and organizations investigating missing person cases as a second loosely coupled system. 1

24 The research will not only reveal the relation between customer knowledge and business processes, it will also aim to determine the relation outside the scope of a conventional, formal organization. It will discuss the unpredictable and dynamic loosely coupled system which emerges and is implemented during the natural disaster emergencies and public safety measures, integrating business process management, knowledge management, disaster management, and public safety. 1.1 Research problem The research problem of the thesis is how to integrate knowledge management with business process management and disaster management to identify the influence of customer knowledge on service provider s business processes during a state of emergency, and the possible social impact of such interaction. The research aims to determine how to methodologically approach the assessment of an emergency response system by applying three management concepts. Firstly, I want to measure customer knowledge related to the emergency response system and identify the cases in which customer knowledge influences the emergency response process. Secondly, I want to detect and map knowledge flows and determine whether it is possible to detect the directions of knowledge flows. And thirdly, I want to verify to what extent the developed research approach can be applied to a new research case which is unrelated to the case used for developing methodology and testing the knowledge-process relations Research problem background Research problems in the field of disaster management, related to response process and knowledge management, are approached in several different contexts and with different data and methods. Based on the state of the art, I present a theoretical research framework which remains open due to growing, developing technologies used for research purposes in the field, and consequently new approaches to the phenomenon. Research was included in the framework and compared with other studies in the field of disaster management. The first set of research includes pure theoretical research works. Geale (2012) used a literature review to examine the field of ethics in the context of disaster management. In addition, Reaves, Termini and Burkle (2014) developed a conceptual renewal of the US Navy 2

25 response in the Pacific Ocean during the mitigation of natural disaster threats, while Roosli and O Keefe (2013) investigated the theoretical aspect of accommodation issues and situation management directly after natural disaster occurrences in Malaysia. Drabek and McEntire (2003) theoretically linked disasters with sociological aspects; Abidi, Leeuw and Klummp (2014) employed a systematic literature review to assess the success of humanitarian supply chain; and Lettieri, Masella and Radaelli (2009) used a theoretical research approach to get an insight into disasters and disaster management in general. The second set consists of empirical research works which can be divided into two subsets: the first examines a broader range in the field of disaster management, while the second focuses on case studies. The subset of general research includes the work of authors (Fekete, Hufschmidt and Kuse, 2014) who were trying to establish the benefits and challenges of resilience and vulnerability in connection with disaster management. Galliara and Prabhawalkar (2012) focused on the role of academic institutions in the field of disaster management, while Nishikawa (2010) partly dedicated his work to the disaster management experiences from Japan in relation with the Yokohama strategy and the Hyogo Framework for Action. Moreover, Wyk and Bean (2011) tried to determine which material and technical resources are of key importance in preventing the consequences of natural disasters. In addition, Jahangiri, Izadhah and Tabibi (2011) conducted an international comparative study focused on disaster management at the level of local communities. Among the published works, there are several case studies where a selected problem is addressed in detail or, alternatively, focus only on a carefully selected geographic area. For example, Aldunce and Leon (2007) were trying to determine the possibilities of improving disaster management in Chile; Bhakta Bandari (2014) conducted a research linked with disaster management and mobilization of social capital in the Kathmandu Valley in Nepal; Crawford, Langston and Bajracharya (2013) used flood events in Queensland to study participative project management for improving the resilience to natural disasters. In the USA, Rademacher (2013) focused on the Sussex County in Delaware in order to identify the means employed by a community for efficient disaster management. A similarly narrow geographical scope of research was selected by Indian researchers Nair and Gupta (2011) they analysed the crisis response plans and disaster management plans in the Sambalpur county in the Indian state of Orissa. 3

26 It was interesting to observe the content intersection of the state of the art, as it soon became apparent that research works keeping up with the development of web technologies and social networks occupied a new niche. This set could be termed the new niche social network dynamics during natural disasters. Gao, Barbier and Goolsby (2011) focused their research on the social network platforms as a tool for active inclusion of the crowd within the natural disaster response process; Alexander (2014) revealed that social media provide a means of communication, information exchange, crowd activation and even control possibilities as an integral part of the geographical information system. Huang, Chan and Hyder (2010) came to a similar conclusion while researching the Web 2.0 and social networks from the perspective of comprehensive solutions for disaster management. Finally, Phan and Airoldi (2015) studied the social dynamics of social network users based on the natural disaster information shared on the social media. Similarly to the research framework of this thesis, many studies in the field of the disaster management which investigate a wider scope should also not be neglected. In business process management, Castillo (2004) linked disaster management with the continuity planning at the Boeing Company. Lumpp, Schneider, Holtz, Mueller, Lenz, Biazetti and Petersen (2008) focused on restoring business processes after a natural disaster and the establishment of work continuity. Hale and Moberg (2005) conducted a research on how to improve the supply chain preparedness to natural disasters; and Schrank et al. (2013) focused on the comprehensive business process restoration of a small company following a natural disaster. Knowledge management as a field of scientific research also found its way into disaster management research. Pathirage, Senevirante, Amaratunga and Haigh (2012) studied the challenges and knowledge factors from the perspective of knowledge management in connection with disaster knowledge. Mendis, Karunananda, Samaratunga and Ratnayake (2009) developed a system for knowledge modelling which can be used for the purposes of disaster management, while Thanurjan and Seneviratne (2009) studied the role of disaster management in restoring accommodation facilities after a natural disaster. Moreover, Sobel and Leeson (2007) focused on the general use of knowledge during disaster management situations. 4

27 1.1.2 Relevance of the research problem Knowledge-oriented processes in unstable organizational environments are a challenging and relevant research problem, especially when it comes to changes of the organizational system. Even tough fields like process management, knowledge management and loosely coupled systems had already been studied in detail, practical implications of these three fields combined have not been studied to a full extent, leaving a significant research lacuna. The relevance of the research problem is two-fold. Firstly, there is no previously developed methodological framework for addressing customer knowledge and emergency response process relations, and for integrating knowledge management, business process management and disaster management within the loosely coupled system. Secondly, there is the relevance of flood-related content. Flood endangered communities in the focal area of research are faced with floods almost every year. Even though the area has historically been exposed to flood events of the Sava River, it is not clear how communities learn from past flood events and response processes and how such knowledge improves their flood resilience and self-reliance. The research will identify possible measures aimed at improving community flood resilience and self-reliance based on learning about and from the flood events and the corresponding response processes. In order to address the research problem, the concept of process-based characteristics of a loosely coupled system will be presented and a conceptual knowledge management solution for such a system will be developed with the ability to analyse knowledge flows, to process patterns and to map system elements and on-going dynamics within the flood response system. With access to data describing the knowledge and process relations during an emergency, and methods used to address the analytical cases in the conventional organizational system, I will create and test a novel research approach to the knowledgeprocess relation within the loosely coupled systems. Relations between process architecture and organizational system s success during flood events are very important factors that can be scaled down to the level of a single activity, entity, knowledge node or link among these elements. In addition, less stringent organizational behaviour should encourage learning processes to develop within an organization that encourage the organizational system to operate better. Yet the reality of such a relation in emergencies remains unclear. 5

28 A knowledge management approach to the loosely coupled system should ensure enough openness for learners to take their own initiative and participate in knowledge-related processes in the system. Only appropriate loosely coupled system tailored methodology would successfully address the uncertainty concerning the changing structure of knowledgebased processes while also providing useful feedback about the state of the process and about knowledge flows. To address the research problem, a framework was developed with the capacity of analysing different types of loosely coupled systems and presenting the relevant scientific achievement. 1.2 Motivation The motivation to address the selected research problem is based on real-life influence of the relation between customer knowledge and business processes in flood endangered communities of the Lower Sava Valley. In the research, customer knowledge is a form of community knowledge that could have a significant capacity to influence flood response processes of responsible responding organizations. Elucidation of such a relation would make it possible to obtain new knowledge and make process-based improvements in the flood resilience and self-reliance of flood endangered communities. Importantly, such insights must be retrieved due to a high flood risk in the area and a lack of efficient approach to natural disaster prevention and education in natural disaster preparedness. Between 2009 and 2015, Slovenia faced several flood events which caused great damage. Surprisingly, floods that should occur only once per century occurred two times in the observed timeframe, accentuating the importance of expanding our understanding of flood endangerment in the local communities, as well as the understanding of the issue from the regional and national perspectives. In November 2014, a special task group led by the Ministry of the Environment and Spatial Planning (Majcen, 2014) prepared a document entitled Flood prevention intervention activity action plan, confirmed also by the Slovenian government. The document includes infrastructure interventions and other actions in the value of 37.5 million EUR, a sign of the government s long-standing negligence towards the management and maintenance of waterways and a lack of flood prevention activities. 6

29 In addition to facts connected with new and unexpected flood impacts, national strategic documents introducing guidelines for disaster prevention system development and national security system development, confirmed by the Slovene parliament (National Assembly of the Republic of Slovenia, 2009, 2010), must be considered. Two important tendencies were identified in the Resolution of the National Program of Disaster Prevention for the Period , and the Resolution of the National Security Strategy (confirmed in 2010), as follows: 1) Reduction and specialization of structures responsible for a response during flooding events. 2) Emphasized awareness of citizens about natural disasters, self-reliance and reduction of flood-connected vulnerability. The research aims to determine how customer knowledge (in the assessed research case: flood-affected communities) affects the service processes of providers on the market (in the assessed research case: flood-responding organizational structures). Motivation for selecting the case is supported by the following arguments: - The middle of the Krško-Brežice Field is the site of the only nuclear power plant in the broader region (including the Balkans, Austria, and Italy), which was designed to be flood safe. Its flood protection consists of an elevated area in the form of an artificially designed plateau with the main infrastructure, and a defence embankment. The existing flood defence would successfully prevent flooding of the power plant but it would also redirect the rising waters towards lower grounds, where settlements, local communities, road infrastructure and even an airport base are located. Considering the case of the highest theoretical Sava River flow (the power plant uses its waters for cooling the reactor), the flooding occurring in the plant s surroundings would have catastrophic dimensions. - The municipalities situated on the Krško-Brežice Field perform the activities connected with reducing and specializing their units differently, consequently creating structural differences in three local systems situated in a relatively small geographical area. - Both municipalities and Administration for Civil Protection and Disaster Relief of the Republic of Slovenia have not developed an efficient and systematic approach to the implementation of identified educational guidelines. Lack of this approach creates a 7

30 situation in which local communities are left to rely on their own knowledge sources and learning methods in order to acquire the necessary knowledge. - The system established as a response to flood events is a loosely coupled system, changing over time due to different knowledge provided by formal structures actively involved in the process, and to the knowledge of local communities. The fact that the system is loosely coupled enables a quick emergence and response, although it also has a downside: it delivers operations which are less transparent and has no efficient success indicators. Our research case is a real, operating system, which emerged with the aim of protecting human lives and property during flood events in an especially vulnerable territory due to close proximity of an active nuclear power plant. Never before has an analysis been conducted about what local communities at risk of flooding are learning and how their flood-related knowledge influences the responding entities and their processes. Understanding this phenomenon importantly contributes to: - process optimization within the flood response entities; - a greater possibility of optimizing the local disaster protection standard operation procedures; - a greater possibility of improving local threat assessment; - an easier planning of new, updated local disaster protection development programs; - easier implementation of additional educational activities in the field of self-reliant protection, outlines in the national strategic documents; - easier initiation of new civil processes for flood prevention and mitigation. 1.3 Hypotheses The research is based on five hypotheses, which include testing the loosely coupled system, its operating processes and its learning processes in order to test the ability of customer learning and emerged knowledge, to influence work and the ongoing processes of the service provider in the loosely coupled system. Hypotheses 1-4 are designed to determine the customer flood endangered households in communities at risk of flood events on a regular basis. In this case, customer providers are the responsible organizations who are allowed and obliged to respond in order to protect the flood-endangered households and communities. The 8

31 fifth hypothesis deals with another case in the field of public safety, where the customer is the victim s family and service providers are all organizations entitled to conduct investigation, and any other organizations and individuals with the capacity to support the investigation. Hypothesis 1 (H1): Loosely coupled systems are built on process architecture that can be defined and assessed. Process architecture in conventional formal organizations usually does not change without a special reason, such as for example improvement of business processes. At the same time, a loosely coupled system is faced with visible and lasting change in its process architecture. Therefore, a feasible approach had to be developed in order to analyse and model the processes running on changeable architecture. Hypothesis 2 (H2): It is possible to map and assess community knowledge. What is community knowledge, how does it emerge, what is its course of interchange and how to trace and record it? These are the questions pertinent to the flood response system which flood endangered communities still have not answered. Considering that knowledge management methodologies are proven to work in other organizational systems, I will adopt and adjust them to be used in the framework of the loosely coupled system. Hypothesis 3 (H3): Community knowledge can influence processes in the loosely coupled system. I believe the influential component in knowledge-business processes relation is present also in the analysed loosely coupled system. The system acts as a dynamic and changeable structure, and it is not yet clear what part in this change is caused due to the influence of knowledge. Based on the insights about customer knowledge from business-oriented and other, conventional formal organizations, I believe community knowledge has an influence on the loosely coupled system. Hypothesis 4 (H4): Community learning in the loosely coupled system is a mutual process. By tracking the learning process in the communities of the loosely coupled system, I will be able to detect and measure the level of mutuality. I believe that different entities involved in the loosely coupled system contribute different knowledge and that the knowledge exchange is a mutual process. 9

32 Hypothesis 5 (H5): Knowledge-based process pattern recognition model can be used for ensuring public safety. Process patterns are widget parts of process architectures. Having the ability to recognize their behaviour based on the learning process and knowledge flow would represent an important contribution to resolving public safety matters. I believe it is possible to apply the results of the research work, based on previous hypotheses, to a public safety case without a direct link to the previous loosely coupled system which will in itself represent a loosely coupled system. 1.4 Methodology The research strategy consists of a methodological design based on qualitative and quantitative approaches to the identified research problem. Structured interviews will be conducted to provide an insight into the domestic flood endangered processes during flood events. Tabular application development (TAD) methodology (Damij, 2000, 2007) will serve as the pilot business process modelling tool. I will use its simplicity and openness in order to capture the process architecture of the loosely coupled system and to identify important work flows. In addition, I will expand the use of the TAD methodology by adding a new knowledgeoriented module. The knowledge module will be used to map the knowledge within the loosely coupled system and to assess the knowledge flows. For the purposes of process simulation, two simulation solutions will be employed. The first, igrafx, will be a professional commercial software for business process simulations, developed by Corel. igrafx introduces the capacity to design an organizational structure with the corresponding departures, human resources, costs, workflows, and to run the simulation for the modelled organization. The final output of the simulation is the report with a statistical overview of process exectuion. I will also design a custom simulation environment in R coding environment to complement the igrafx simulation with inclusion of the data which could not be used as the input for igrafx. The custom simulation will simulate water levels of the rivers, geographical variables such as altitude, longitude and latitude, knowledge exchange, human behaviour, disaster response processes, and communication within the loosely coupled system. 10

33 A new process optimization algorithm will also be designed by using igrafx as the platform for running the decision flow. The process optimization algorithm will predict how future knowledge would affect the flood response process. Statistical methods used will include Student's t-test, multivariate analysis of variance and other appropriate analytical methods to analyse the obtained numerical and categorical data. In addition, datamining algorithms will also be used for data analysis and implemented in software solutions such as Weka or RapidMiner. In order to obtain the data regarding the process execution, reactions and behaviour of flood endangered households, and data regarding the missing person investigation process, several semi-structured interviews will be conducted. By considering the table for determining the sample size, developed by Krejcie and Morgan (1970), sample size with the population N=167 should be somewhere between 113 and 118. A preliminary interview session was conducted with the sample size of 22, followed by semi-structured interview sessions with a sample size of 56. Two criteria groups were used to select the sample. The first criteria group required that respondents had to be actively present on site during at least two flood events, that their households must have been directly or indirectly endangered by the flooding water, and, last but not least, that they had the ability to reach an information source during the flood event. The second criteria group defined the respondents profile and consequently the sample structure, based on harmonization of distribution and standard deviation of two attributes: municipality and threat source. This measure enabled the sample to be more representative from the perspective of municipalities where they come from and the manner in which their households were endangered by the flood events. Even though the preliminary sample size was increased by more than 50%, the sample size recommended by Krejcie and Morgan (1970) was still not reached. There are two main reasons for difficult recruitment of additional respondents. The first reason is that some of the residents who were invited to participate in the research expected direct intervention results. One of their first questions was whether I will be able to practically improve the short-term flood protection of their homes, their belongings, and their land. Due to the fact that the research itself will have no implementable short-term impacts on the flood protection system in flood endangered local communities, many invited persons declined to participate in the research. As a result, the second reason was to adjust the number of respondents in 11

34 accordance with the predefined standard deviation and distribution of attributes municipality and threat source. The aim of the interviews was to identify how participants perceive their own learning process during flood events, the knowledge they gain, and what are the most important pillars of flood response. The semi-structured interviews were not used as a method for evaluating the effectiveness of learning; rather, a fuzzy set approach was employed for gauging effectiveness. Finally, I was also interested in the daily lives of respondents during flood events. Because the number of interviewed participants was lower than suggested, the methodological framework also included four semi-structured interviews with representatives of organizations responsible for overseeing the practical education and knowledge sharing among local communities in the broader region. These organizations include: the Civil Protection Department of the Brežice municipality (Mayor s counsellor and deputy commander), the regional Brežice office of the Administration for Civil Protection and Disaster Relief (head of the office), Department for Protection, Rescue and Civil Defence of Ljubljana municipality (higher counsellor), Administration for Civil Protection and Disaster Relief (Deputy Director General). The local civil protection or civil defence department is an integral part of the local government, and as such subject to local standard operation procedures and other local acts confirmed by the local town council. What is more, its operations must be in accordance with the national legislation, and with regional and national standard operation procedures. The regional office of the Administration for Civil Protection and Disaster Relief is the regional authority subordinate to the national Civil Protection headquarters, governmental, regional and national standard operating procedures and other acts confirmed by the Administration for Civil Protection and Disaster Relief, the Ministry of Defence, or the government.»administration of the Republic of Slovenia for Civil Protection and Disaster Relief is a constituent body of the Ministry of Defence. It performs administrative and professional protection, rescue and relief tasks as well as other tasks regarding protection against natural and other disasters. Administration is divided into six internal organizational units (four sectors and two services) based in Ljubljana as well as 13 other branches operating 12

35 throughout Slovenia. Within each branch there is a regional notification centre that performs a 24-hour duty service«(administration of the Republic of Slovenia for Civil Protection and Disaster Relief, 2015). A semi-structured interview was used as the main qualitative methodological approach in order to gather part of the data because there was no previous reported knowledge of how and to what extent the residents and communities in the flood endangered areas learned. Never before has a similar research with the same focus been conducted in the area. Therefore, the semi-structured interviews allowed the development of a theoretical contentment framework and identification of the unknowns, explained in more detail in the interview itself. The predefined theoretical content was filled and complemented on the basis of participants answers, as described by Galetta (2013). The respondents were led through the content so that they were able to provide an insight into their own response during flood events and their learning process. Another argument to keep the semi-structured interview as part of the methodological framework was also the diversity of respondents perception of the situation in which they were facing flood events (location, type of flooding threat, level of flooding threat, demographic, social and economic characteristics). The goal was to have the smallest possible influence on respondents answers and to expose them to the smallest possible number of content constrains. This part of the research included 56 participants with a mean age of years and a standard deviation of years. Additional information on the population and the sample is shown in Table 1.1. Table 1.1: Sample presentation Categorical values Brežice 1 Municipality Kostanjevica na Krki 2 Krško 3 Torrent 1 Krka river 2 Threat Meteorological source water 3 Ground water 4 Sava River 5 Diversity of respondents (municipality + flood threat) Source: Agrež, own research (2015) Population (167) Sample (56) standard mean deviation value mean value standard deviation

36 Table 1.1 shows the categorical values for the attributes municipality (1 4) and threat source (1 5). Further, the table includes mean values and standard deviations of these attributes, calculated for the population and for the sample. At the bottom of the table is also the difference between respondents according to the population and the sample. The diversity measure was calculated by counting the number of combinations between the municipality and threat source variables (1 15). 1.5 Research objectives The aim of this research is to determine how customer knowledge relates to and influences business processes in an unstable organizational environment, i.e. a loosely coupled system, and how this correlation impacts the surrounding social environment. In order to do so, a practical and applicable research framework will have to be designed to reach the objectives outlined below. This framework must be able to integrate a knowledge map, follow the evolution of a knowledge network, statistically analyse the emerging knowledge-based process patterns, and design to-be simulations of the system. Firstly, it must be determined whether conventional business process management methodologies can be applied to unconventional organizational cases such as a loosely coupled organizational system. Next, it has to be established which of the existing methodologies is most appropriate and how to adjust it for the purposes of assessing, modelling and simulating a loosely coupled system. Secondly, it has to be defined how to assess, track and map the knowledge accumulated in communities which are part of the loosely coupled system. A community-learning tool has to be developed and learning criteria defined in order to obtain an insight into the structure of community knowledge and its allocation. Thirdly, I want to identify whether knowledge in the community represents a measure that could influence the ongoing processes within the loosely coupled system. To do so, I must assess the current (as-is) state of the process and consider the detected and evaluated knowledge. To reach my objective, I will project the knowledge that could emerge in the communities in order to assess this new, future knowledge and trace its possible influences. Another important aspect of the knowledge-process relation is to reach an understanding of whether learning in the loosely coupled process is a mutual process including several 14

37 knowledge sharing entities and creating a common knowledge flow. It is important to understand where to launch a learning process regarding the mutuality aspect to be able to better explain how knowledge influences the ongoing processes in the loosely coupled system. Finally, I want to verify whether it is possible to apply a designed, single case tested methodological frame to another case related to public safety, but not directly linked to the original field case. The appropriate case will have to be identified and all the necessary data retrieved to do so. With successful verification, I will be able to expand my research from a narrowly focused case to gain a wider perspective. 1.6 Scientific contribution Integrating business process management with knowledge management and applying this integration to a loosely coupled system with the aim of observing customer-provider relations based on customer knowledge is a novel research approach, present in the state of the art only partially. Selecting the flood response system and the missing person case investigation as examples of a real-life loosely coupled system linked the primary two fields of management with disaster management, creating an even narrower research niche from the perspective of knowledge based customer-provider relation. To sum up, this work will specifically contribute to: - revealing the influence of customer knowledge on business processes during emergencies, - extending the conventional business process methodology with data mining and different statistical methods, - complementing process and knowledge related data with the meteorological, hydrological and geographical data, - complementing business process modelling and simulation solutions with knowledge mapping and knowledge flow simulations, - providing a solution for simulating future knowledge and its influences, - expanding the use of developed methodology to a new, independent public safety case. 15

38 In the trinity of business process management, knowledge management and disaster management, the research problem of how customer knowledge influences business processes of a service provider and consequently triggers social influence, remains unanswered, while the prediction of how future knowledge might affect business processes seems like a completely new idea in knowledge observation. 1.7 Structure of the thesis The research work consists of four major sections: I first introduce the general research problem, the motivation for the research, a general methodological approach and specific objectives of the research. Next is a presentation of the key concepts and literature review, where I introduce state of the art in the general theoretical fields the research is based on. The literature review is divided into three subsections addressing the areas of system and organization, business process management, and knowledge management as the three main theoretical pillars of the research. The following section is empirical research, beginning with the background introduction of the research cases and data used during verification of the first four hypotheses. Further, I present the methodology and results of the first four hypotheses, and, finally, due to a separate research case, I present the case background, data, methodology and results related to hypothesis number five. In the empirical part of the research, a real-life loosely coupled system is transformed into a custom designed simulation with aspects of all three fields: business process management, knowledge management, and disaster management. The simulation design offered an opportunity to analyse the process-based validity of the loosely coupled system, which is also presented. Further, community knowledge was mapped and assessed it represents customer knowledge in the flood response system. I was able to establish what kind of knowledge is present in communities and how the detected knowledge is obtained. The third part of the thesis aims to determine whether community knowledge has the power to affect the flood response process by any traceable field or other influence. In the fourth part, the research focuses on whether community knowledge, as the potential flood response process influence, emerges trough a mutual learning process. The learning mutuality between customers (flood endangered entities) and service providers (flood responding organizations) was assessed. In the fifth and final part of the research, the methodology designed during the previous phase 16

39 was adopted and applied to the case from the perspective of public safety in order to verify whether the developed methodological approach is applicable also for analysing knowledgebased customer-provider relations in cases other than flood events. In the final section, the reached objectives, contribution of research, its practical application and research limitations are described, followed in the end with the possible improvements and further research proposals. 2 KEY CONCEPTS AND LITERATURE REVIEW In this chapter, the established theoretical concept and literature review are presented, beginning with the definitions of the system and organization, identification of loosely coupled systems and a comparison of looseness, sustainability and continuity. I continue with the business process management subsection, where I introduce the business process architecture, business process improvement, business process modelling, and business process simulations. The final subsection is dedicated to knowledge management. In this subsection, organizational knowledge, tacit, explicit and experiential knowledge, experiential learning, learning in the loosely coupled processes and open knowledge management approach in a loosely coupled system are discussed. 2.1 System and organization What is a system and what does it represents? Leveson (2013) argues that a system could be defined as a set of components working together as a whole in order to achieve a common goal. These components establish direct or indirect connections among each other, where one system could easily become part of a bigger system through the process of connecting several components or subsystems. Levchenko and Kotolupov (2009) clearly defined the scope of a system as spanning from single cell organisms, where organelles take over the role of the structural subsystems, followed by multi-cell organism, where organs and cells assume the role of subsystems. They also mention ecosystems, where subsystems are represented by organisms, species and smaller ecosystems. Finally, they present the biosphere, where giogeocenosises serve as a subsystem. 17

40 On the other hand, Handy (1993) describes an early interpretation of the term organization as machines with human parts which can be tailored, designed according to the requirements, controlled, accelerated and slowed down. All these operations could be managed by an external agent. Handy (1993) continues by arguing that the modern interpretation of the term organization is based on a collective of productive, useful people, while approximately three decades earlier, Evans and Lynch (1973) connected the modern concept of organization with efficiency and rationality, which is also highlighted by Thompson (2011). Recently, Shafritz, Ott and Jang (2015) included contemporary organizations among clearly defined structures which can be presented with a transparent internal structure and are also articulated and formalized. That organizations should go hand in hand with formality is a concept also supported by Miles (2012), who stressed that an organization cannot be equated with a randomly gathered group of people, but must rather be a consciously, formally established entity based on the aim of reaching the common goal which individuals could not reach by their independent actions. Thus, it becomes clear that organization is a formalized, transparent system with a clearly defined structure, processes and goals, while a system could be also a structure of a less formal, yet still systematic, but at the same time less predictive nature (i.e.: a biosphere). When the characteristics of the loosely coupled and tight systems are compared with the characteristics of the organization, it is easy to determine that an organization fits the characteristics of a tight system more Loose coupling Beekun and Gick (2001) determined that coupling usually means establishing a relationship among the elements of at least two variables, where loose coupling emphasizes general characteristics of such a relationship the level of loose coupling or the tightness of the relationship. The differences between tight and loosely coupled systems are shown in Table 2.1. Table 2.1: Differences between loosely coupled and tight systems Coupling components Loosely coupled systems Tight systems Coupling elements Subunits are making couples with other subunits (Snook, 2000), systems are making couples with other systems 18

41 (Mayer and Whittingtion, 1999), success indicators are making couples with determinations and goals (Johnsen, 1999). Coupling domains Identified coupling domains are: work-related communication domain (Grabowski and Roberts, 1998), non-work related communication domain (Covaleski and Dirsmith, 1983), workflow domain (Ibarra, 1993), bureaucracy domain, including support activities (Kerwood, 1995), resource exchange domain (Kerwood, 1995), structural domain with structural changes and forms of the system (Sanchez, 1997), and social domain with informal communication and social activities (Kerwood, 1995). Usual characteristics: spontaneity and nonplanning. Usual characteristics: previous planning, deliberate and systematic implementation. Coupling dimensions Coupling dimensions indicate the quality of relations among the coupled elements (Weick, 1982). - coupling strength - directness Low frequency of the interactions among the elements, low intensity of the interactions, short lasting interactions. Lower level of directness (Weick, 1982, Svetlik, Ilič and Sadar, 2006), higher possibility of communication noise, weak communication connections. High frequency of the interactions among the elements, high intensity of the interactions, long lasting interactions. Higher level of directness, lower possibility of communication noise, strong communication connections. - consistency High diversity of external Lower diversity of the 19

42 - dependence stimuli (Orton and Weick, 1990), ability of diverse adaptation (Perrow, 1984). Higher autonomy of the systems elements (Weick, 1976). reactions to the external stimuli, low ability of diverse adaptation. Higher dependence of the systems elements. Coupling mechanisms A coupling mechanism describes the processes which enable consistent and concomitant operations of the systems elements (Beekun and Glick, 2001). - differentiation (segmentation of the High level of differentiation system into (Lawrence and Lorsch, Low level of differentiation. subsystems), 1967). described by Lawrence and Lorsch (1967) - integration (possibility of achieving unity of Low level of integration High level of integration several subsystems), ability (Perrow, 1984). ability (Perrow, 1984). described by Lawrence and Lorsch (1967) Adapted according to: Snook (2000); Mayer and Whittingtion (1999); Johnsen (1999); Grabowski and Roberts (1998); Covaleski and Dirsmith (1983); Ibarra (1993); Kerwood (1995); Sanchez (1997); Weick (1982); Svetlik, Ilič and Sadar (2006); Orton and Weick (1990); Perrow (1984); Weick (1976); Beekun and Glick (2001); Lawrence and Lorsch (1967). Table 2.1 provides a clear insight into the differences between conventional, tight systems and loosely coupled systems, comparing both types of systems from the perspective of coupling 20

43 elements, domains, dimensions, strength, directness, consistency, dependence and coupling mechanisms Looseness and sustainability, looseness and continuity In order to determine the connections between the looseness of a system and the time dimension of its operation, it must first be established what the time dimension in the assessed case represents. The term sustainability often seems to be connected with the concept of sustainable development. Smith (2011) discusses organizational sustainability, connecting it with corporational social responsibility, while Craig and Allen (2012) equate sustainability within the organizational system with the improvement of the ecological conditions around us. Similarly, Berns et al. (2009) connect sustainability of the organizational system with the ecological influence, consequently referring to different economic, social and personal influences. Frandsen et al. (2013) even promote the introduction of a loose framework for providing sustainability as the concept for achieving ecological orientation of an organization, due to the fact a loose framework should encourage a more relaxed and productive interaction of all those encouraging and managing the transition to a sustainable business. A literature review emphasized the term continuity, which is more closely described by the time dimension of the system s operation in terms of identified needs than the term sustainability. Lindström et al. (2010) relate the continuity of the work within the organization system with the ability to handle any kind of situation and consequently to keep the system operating. Kolowitz et al. (2012) define the term continuous workflow as the ability to provide uninterrupted service and put it in close connection with business continuity. Moreover, Bakar et al. (2015) developed a conceptual frame for managing business continuity, which led them to the conclusion that managing business continuity is of high importance for decreasing operation risks and the risk of failure of important business functions. In his case study of non-profit organizations, Mukheerje (2014) related continuity with constant growth of an organization. According to Perrow (1984), a comparison between loosely coupled systems with tight systems revealed that the former have a higher ability to limit the damage possibly arising from negative events, which is closely related with the modern concept of business continuity. Moreover, in their literature review related to the concept of continuity, Spender and Grinyer (1995) connected the term continuity with continuous renewal of the organizational system. 21

44 They argued that there is a visible difference between continuous renewal and organizational change, which they related to the marginal change of the system s components. Therefore, the term continuity can be connected with the existence of a system. Both Spender and Grinyer (1995) and Orton and Weick (1990) relate the continuity of a loosely coupled system with the phenomenon simply termed glue. Glue is a phenomenon which enables subsystems to be loosely connected into the whole system, but the dilemma arises of how it is possible to set up a system when there is a lack of communication between separate subsystems right from the start (Spender and Grinyer, 1995). The common values of the subsystems within the system offer a solution for the lack of communication. As a result, the system cannot exist until value-sharing environment is established. According to Jaakson (2010), values of an organization influence different fields of the organizational system, including its goals, behaviour and characteristics of systems stakeholders. Sousa and Porto (2015) define organizational values as the leading principles of organizational system existence, while Benneth et al. (2009) connect them closely with knowledge created within the organizational systems in a directed manner, depending on the existence of values. The system assessed in this thesis is a loosely coupled system, and when observing its continuity, I notice that it emerges according to the needs stemming from its surroundings and that it is not actively present all the time. On the other hand, its subsystems do not have the same characteristics. In Table 2.2, the aforementioned subsystems are compared in order to determine the common characteristics influencing the examined loosely coupled system. A theoretical comparison of the practical subsystems of a loosely coupled flood response system revealed that only a flood threatened community is an example of a loose system, the rest are all clear examples of tight organizational systems. Table 2.2: Comparison of the subsystems Subsystem Loosely coupled / tight Workflow continuity Values (in relation to the entire system) Present knowledge Type of subsystem Role in the entire system Flood endangered communities Loosely coupled Continuous work flow Safety Safety breach experience Social network Requesters Local civil Tight Semicontinuous Readiness to Understanding the local flood A system established by the Responders 22

45 protection work flow help, legal duty danger characteristics, understanding proper actions and use of equipment city council decree, of non-profit nature and operating mostly on a voluntary basis Understanding the local flood Local firefighting association Tight Semicontinuous work flow Readiness to help, legal duty danger characteristics, understanding proper actions Non-profit organization, operating mostly on a voluntary basis Responders and use of equipment Distress call centre / regional office of Administration for Civil Protection and Disaster Relief Tight Continuous work flow Legal duty Understanding the local flood danger characteristics, understanding proper actions and use of equipment Part of the Ministry of Defence Responders Road company Tight Continuous work flow Readiness to provide a service, additional income source Understanding proper actions and use of equipment Company, profit organization Responders Readiness to Civil engineering company Tight Continuous work flow provide a service, additional Understanding proper actions Company, profit organization Responders income source Communal company Tight Continuous work flow Readiness to provide a service, additional income source Understanding proper actions and use of equipment Company, prot organization Responders Source: Agrež, own research (2015) 23

46 When comparing from the perspective of continuity, it becomes clear that the local civil protection and local firefighting association are the only ones without a continuous workflow. Both operate on the basis of semi-continuous workflow, meaning that they provide continuity only for the minimum amount of their capacity, while the rest, including personnel and equipment, are activated only when necessary and for a limited amount of time. Even though the operation of these subsystems is not fully continuous, their role during the emergence of a flood protection loosely coupled system is of high importance. The flood endangered community as the main pillar of the system is continuously present as a social network, with no significant breaks. Nevertheless, the flood threat as its main characteristic during flood events is not present continuously, but appears only a few times annually, under specific meteorological and hydrological conditions. The Regional Distress Centre, operating under the regional office of the Administration for Civil Protection and Disaster Relief, presents the purest continuous workflow. Among all included subsystems, its operations are always being executed, without interruptions. In fact, it can be said that the Centre s operation is set up and maintained to prevent failure of its workflow, regardless of the circumstances. All three companies within the system operate with a continuous workflow, but differ from the others by the values which represent their links to the system. An overview of the values which link the subsystems into a whole system reveals that the main value among all activation circumstances is in close connection with safety. The connection appears in request subsystems as well as responding subsystems. Communities desire to remain safe, while responding organizations wish to provide the safety or at least to restore it when necessary. Companies active in the system are the contractors, outsourced by the municipality or the Administration for Civil Protection and Disaster Relief. Consequently, companies provide their services because they gain financial income, while field interventions executed by the local civil protection and firefighting brigades are conducted on a voluntary basis. Conversely, the distress call centre functions as a fully professional, non-profit organization. The Ministry of Defence covers all its costs connected with field intervention. Safety either as a basic value (Earnest, 2000) or one of the presented subsystems values (Cooper, 2001) is, when taking into account also the work of Sneddon et al. (2006) on systems facing a high possibility of safety risk, in close connection with situational awareness. Situations in which hydro-meteorological conditions cause a direct flood threat are not continuous situations. Such situations appear only a few times per year in the form of isolated 24

47 flood events. Therefore, it can be claimed that such dynamics influence also the perception of safety in connection with a direct flood threat. The whole flood response system in fact only emerges when situational awareness underlines safety as a value endangered by the elements (weather conditions and rising waters). The system conducts a focused operation as long as a direct flood threat exists and the danger for communities persists, which is called the response phase (Zukowski, 2014). The system can extend its operational timeframe to include the time after the direct threat in order to facilitate the overcoming of damage caused by the flood event. This phase is called the recovery phase (Beggan, 2011). The system s operation does not remain active because the responding subsystems gradually lose their active role within the system and return to the state before the flood response system emerged. It is easy to see that the specific case analysed here also reveals its specific characteristics. The flood response system is limited from the perspective of its lifetime only to the response process, with no early, systematic preparation and without later long-term relief and mitigative activities. This is the mapping of reality, and possible future change cannot be completely excluded. 2.2 Business process management A brief overview shows that the processes of evaluation and improvement have been present within organizational systems since the early twentieth century (Paim, 2008), when Taylor (1914) began studying different aspects of workers characteristics. Nowadays, the processoriented approach is used for planning, evaluation, optimization, and establishment, and is applied in several fields such as economy, production, military, industry, education, logistics, etc. As Gulledge et al. see it, business process management allows the implementation of any modern system (2002), helping to produce more timely and accurate information leading to better decision making, to improve transparency leading to better risk management, and to improve auditing operations leading to lower compliance costs (Cerniauskas and Tarantino, 2009). The definition of the business process is termed in such a way to enable the translation of any organization, enterprise, corporation, or just a simple set of activities into a business process flow. This universal framework can be used by following the described steps, the first 25

48 one covering the need to define how a business process is set up within an organizational environment. When looking for conclusions drawn from the existing business process, a need emerges to recognize the process flow dynamics and widgets that constitute it. This stage includes the identification and thorough analysis of all activities and tasks within an organization (Bucher and Winter, 2009). There are many different ways of fragmenting and analysing the connections among them: they could be viewed from different perspectives, resulting in the allocation of different levels of importance to different types of relationships (Bititci and Muir, 1997). Kettinger et al. (1997) use the term diagnose when describing business process identification, emphasizing documentation of the current process and sub-processes in terms of process attributes such as activities, resources, communication, roles, IT, and cost. In the 19 th century, Frederick Winslow Taylor was among the first to study different aspects of workers characteristics in order to improve labour productivity. His management model was based on the practical specialization of work activities (Taylor, 1914). From this period, Henry Fayol, who claimed that management is made of actions close to the majority of organizations, should also not be neglected; his thesis was confirmed trough practical implementation (Brunsson, 2008). At the same time, Gantt designed a tool later named the Gantt chart which could be seen as the very first process model ever used (Gantt, 1913). Taylor s work was followed and expanded by Henry Ford who developed a concept based on the principles of product standardization, the use of specialized equipment, and a reduction of workers specialization for the purpose of production (Tolliday and Zeitlin, 1987) and by Frank Bunker Gilbreth Sr. who introduced Therblig s elemental work motions with the aim of improving work results through constant improvement of the working conditions (Nanda, 2006). According to George (1968), he was the first person to introduce a flowchart and a Functional Flow Block Diagram (FFBD), and also the first to implement the basic cyclic approach, a novelty in total quality management that was subsequently upgraded by Shewhart in 1945 and Juan in 1951 (Reeves and Bednar, 1994). The 1960s brought the Petri net process modelling technique, which became popular in modelling and analysing different kinds of processes, from protocols to business processes (van der Aalst, 1998). In 1970s, IDEF0 was developed by the US military for conducting analyses and estimations, and later evolved in its higher versions (Grover and Kettinger, 2000). Post-Fordism brought fundamental changes, comparing it to earlier Ford s philosophy: 26

49 production line work got replaced by an approach that supported learning, motivation and the provision of know-how to be able to guarantee dynamic answers to market demands (McDowell, 1991). In the early 1980s came the first mention of the theory of constraints which brought new ideas about business planning that attracted many executives and production planners (Rand, 2000), as well as total quality management (Powel, 1995). During the same decade, the Toyota production system (TPS) led Toyota to become one of the largest automobile producers in the world (Fujimoto, 1999), and a few years later, Porter (1985) labelled 1985 as the year of value chain as a new way of presenting organizations activities. The 1990s began with the Rummler-Brache methodology, described in their book Improving Performance (Rummler and Brache, 1995), and business process improvement (also known as business process redesign), described in James Harrington s Business Process Improvement: The Breakthrough Strategy for Total Quality, Productivity, and Competitiveness (Harrington, 1991). This is also the period of business process reengineering (also known as business process innovation), defined as a global innovation and thorough change of business processes to achieve important improvements in performance measures such as cost, quality, service and speed of production (Hammer and Champy 2009). What is more, enterprise resource planning was deployed and promoted among executives officers (Umble et al., 2003) and lean production, derived from TPS and influenced by Taylorism and Fordism, began its successful march (Womack and Jones, 2003). Gupta (2009) propragated the importance of process-based link with knowledge, now known as a learning organization, together with Nonaka (1991), who advocated understanding knowledge within the frames of knowledge management. Capability maturity model was developed and included a set of criteria that can serve to improve the organizational development processes (Paulk et al., 1993). Benchmarking was first described as a business tool in Boxwell s Benchmarking for Competitive Advantage, and soon after that came a notable Six sigma breakthrough in General Electric (Eckes, 2001) together with workflow management and its main goal: a systematic process oriented approach (van der Aalst, 1998). The 21st century saw an expansion of supply chain management, connecting organizations that ran operational and managerial valueadding processes (Lambert and Cooper, 2000). The capability maturity model evolved within the business process maturity model and is still the subject of current research (Hüffner, 2004), while process architecture based management 27

50 got introduced through the service oriented architecture, revealing the component-based process view (Bieberstein et al., 2008). Process automation appeared together with enterprise application integration. Both are symbiotically incorporated in process modelling (Linthicum, 2003), and business process collaboration became a very important part of the current global business development (Gong et al., 2006). Business process mining appeared as a useful tool for process identification and definition, with the ability to extract the knowledge about processes from their transactional log files (van der Aalst et al., 2003). In 2011, the business process model and notation 2.0 were released, replacing the first generation of BPMN and becoming a standardized notation for writing business processes (Allweyer, 2010). With the rapid development of information technology, influences can also be seen in IS, with its business part not excluded. Real-time business intelligence is becoming the decision driver, disabling process slowdown, and supporting transformation from short-term process management (Plenkiewicz, 2010) first to day-to-day process management (Cohen, 2010) and finally to real-time process management (Smith et al., 2003). It is not hard to establish that from the early days of scientific advances until today, the vertical structure has been developed in the following way: first came process modelling, then process improvement. Each of these models contains many different approaches that use a specific set of methods. From the horizontal point of view, such structure is intertwined with similar western or eastern influences necessary for establishing connections, and enables a synergy among the approaches and methodologies. Such synergies produce market products, suits that became known as business process management systems (BPMS) which contain different solutions for specific demands, allowing tailored access to optimal results. BPMS can be defined as the latest generation in the process management concept after a century of development, testing and expanding the process management structure, modern systems are trying to merge it and use its best aspects: the perfect elements that will yield the required solutions. Nowadays, the colourful family of business process management approaches provides users almost with a tailored solution for their managerial needs. From the simplest modelling, sophisticated simulations, and statistical analysis of the ongoing processes, to the predictions and cost benefit projections of the process modifications: business process management is a necessity in every organization that wants to operate transparently and have full control of its operations. 28

51 2.2.1 Business process architecture Business process simulation combines visualization with statistical and mathematical presentations of its dynamics and attributes, including costs, cycle time, serviceability and resource utilization (Paul et al., 1999; Levas et al., 1995; cited in Serrano and Hengst, 2005). This is why they are used, as argued by Helquist et al. (2012), for systematic examination of potential risk within the process architecture of the optimized process model. Only changes that affect the core standard repeated processes of a business model constitute a change in the business model (Cavalante et al., 2011). Simulation can be used for predicting the results of change business process innovation and business process improvement. The main distinction between approaches lies in the fundamental logic of conducting optimization. To simulate business process innovation, I actually set up new process dynamics with familiar context, inputs and outputs, whereas for process improvement, I used the same process flow and added the necessary adjustments to meet the optimization aim. By establishing and defining a complex process, the modelling itself is not able to offer the desired precision of information that would show the direction of approaching process modification. This is why process identification and modelling provide insights into the business process within the scope of process review. Simulation, on the other hand, goes one step further towards dynamic change: it enables innovation or improvement. Three intermediate stages of addressing business process management can be adopted between business process modelling and simulation: architecture, choreography and orchestration. Barros and Julio (2011) describe business process architecture as a design of the relationships that coordinate all the components of an architecture and make them perform as a system. For such design, it is very useful to have a general architecture model that explicitly provides the relationship processes and other elements it should have (Barros and Julio, 2011). Business process architecture supports the establishment of design that includes process patterns required to reach the aimed output. It can be applied in different parts of a business process and used as a common solution for similar problems that lower the optimization level of the process. Strnadl (2006) argues that a process diagram includes the features of an organizational chart of how an enterprise is designed and allows analysis, changes in design, managerial improvement and optimization of the dynamic structure of a business, which constitutes a basic ingredient for achieving speed and agility (Strnadl, 2006). 29

52 Within this framework, he finds sufficient space to initiate the process of architecture solutions. Just as different product architectures can deliver varied product capabilities and levels of effectiveness, alternative process architectures have a different cost, duration, and risk characteristics. Much like a product can be improved through architectural innovation, process improvement includes architecting an efficient and predictable process (Browning and Eppinger, 2002). Another approach toward business process design optimization is choreography. Business process choreography is a formal methodology for representing interoperability patterns between two business processes and for automating the patterns systematically (Jung et al., 2004). In fact, business process architecture and choreography indicate possibilities of complementation. Architecture provides recognized process patterns that are applied for specific inputs and outputs, while choreography uses process interoperability to connect these fragmented patterns from separate processes into a common, sufficient and effective business process. Similar to choreography is the process of orchestration, referring to an executable part of an inter-organizational process that is provided by one party. Although it contributes to successful execution of a global choreography, the state of the orchestration process is controlled locally (Mendling and Hafner, 2008). Van der Aalst and Weske (2003) (cited in Mendling and Hafner, 2008), argue that a common approach in the process field is to reach synchronization among multiple parties in global choreography; in this research, however, local orchestrations can be used. Business process architecture offers an interesting analytical view on the business processes wider than the activity level and yet more specific than the process itself. Built out of process patterns, it provides the opportunity to identify and monitor similarities and deviations of the workflows within the business processes Business process improvement Very important parts of business process optimization are process innovation and process improvement. They both aim towards a greater process optimization, but employ completely different approaches to reach this aim. When modelling a business process, it soon becomes evident if activity flow is clear and logical or if there are any obvious unnecessary nestings, bottlenecks or cost sources making the process less optimal. Simulation takes it one step further by revealing how the implementation of a selected process change could contribute to 30

53 its optimization. If there are indications that small changes could contribute to a better result, process improvement can be employed. In reality, enterprises have limited resources of investments and their strategies are focused. Therefore, the improvement boundaries, goal and direction of the process should be redefined depending on the situation of the enterprise (Ma et al., 2012). This means that the core process remains the same with additional improvements. However, when deep, holistic change is required, process improvement will not be sufficient. In this case, business process innovation is the solution. In this research, the complete process has been innovatively redesigned, and new orchestration rules and choreography junctions were set. This solution must be strategically verified and carefully planned because it incorporates critical, meaningful changes for the whole business system and will not influence only the local aspect of the process. Business process redesign focuses on the fundamental rethinking of the business process, ignoring organizational boundaries (van der Aalst and He, 1996), which is why a prior simulation of such a method is crucial to avoid radical mistakes that can produce irreparable damage in the addressed business system Business process modelling When describing a business process, it must be supported with a presentation model that assumes two important dimensions: to capture the existing process by structurally representing their activities and related elements, and to represent new processes in order to evaluate their performance (Lin et al., 2002). The ongoing and strengthened interest in modelling for business process management has given rise to a wide range of modelling techniques (Recker et al. 2006), such as flowchart diagram, Gantt chart diagram, UML, BPMN, IDEF modelling family, second generation of BPMN, Petri nets, role activity diagrams, and others. All of them are based on primary dynamic, such as the transition between two activities, functions, roles, etc., that is supported by secondary dynamic which defines additional changes besides the primary transitions. Recker et al. (2006) divide modelling approaches into two groups. The first group supports intuitive modelling which aims to capture, understand and explain the process model. Modelling with such approaches is not complex by default. Some methods, such as workflow diagram or Gantt chart diagram, even do not allow high levels of complexity integration, 31

54 because they do not hold the appropriate mechanisms. On the other hand, methods like Petri nets or the business process modeling language allow a higher level of complexity but only proportional to the complexity of the real process. The second group, however, is based on mathematical laws and programed algorithms, and is usually used for process analysis, execution and simulation. Modelling with the latter group enables a complex dynamic including functions, relations and statistically based features. Next, the modelling approaches of both previously defined groups will be presented. Group 1 methods share the common aim of presenting processes in a simple, easy and understandable manner. As a result of their simplicity, they represent a useful tool for any kind of organization which tries to capture and analyse its own processes, but at the same time allows a detailed study of patterns and relations. Methods from Group 2 use a graphic presentation of the addressed process as a visual aspect of the coded dynamic running in the background. This allows the running of algorithmic and numerical simulations, creating detailed events and capturing statistical data for further analysis. IDEF stands for Integrated Computer-Aided Manufacturing Definition Language, which is a list of rules that is used for supporting process re-engineering, developed by the US military to ease process automation. In defence departments, models have been developed with the primary aim of conducting analyses and defining improvement possibilities (Kappes 1997). Process modelling with IDEF0 rules is a method based on a function which is defined as a set of activities that, by means of some mechanism, transforms the inputs into outputs (Kim et al. 2002). The function creates outputs through the provided control mechanisms. It uses the ICOM (Input, Control, Output, and Mechanism) concept as the main process presentation and takes information or physical objects as input. ICOM boxes are connected by arrows; outputs of one box can be inputs or controls of other boxes (Kacprzak and Kaczmarczyk 2006). IDEF0 modelling is simple and its notation is standard enough that it can be easily transformed into a normalized database (Hastings and Funk 2008). Such a modelling approach does not offer the ability to observe and manipulate time frames, possible mistakes and compliance, but it does provide the possibility of highlighting model incorrectness when it originates in the failure to follow schematic modelling rules (Kacprzak and Kaczmarczyk, 2006). IDEF1 uses the information modelling approach that defines information existing within the observed system. An IDEF1 diagram represents entities that belong to classes, have attributes 32

55 and are linked together via relations (Kokolakis et al., 2000). Bal (1998) developed the definition even further, describing the approach as a method that can conduct an analysis of the existing organizational system, define the problems and point out information requirements. IDEF1 has gained its popularity due to the ability to support communication among the entities in the organization, consequently reflected in the expansion of tools that supported the process (Mayer et al., 1992) of the entity-relation diagram modelling (Lin, 2002). It incorporates the necessary graphics, text, and forms to inject an organized discipline into the process. It provides the measurement and control of the progressive development of the model through the routine of the modelling discipline (Mayer et al., 1992), together with the ability to provide assistance for the discovering, organization and documentation activities of physical and conceptual entities that are present in real environments (Linghzhi, 1996). IDEF1 consists of a powerful set of rules that enable the identification of different objects that exist within the system, their relations, information that exists about the objects, and data structure that forms the set. The IDEF1X technique lends itself to the design and implementation of data models (Ma et al., 2002). According to the definition proposed by Kacprzak and Kaczmarczyk (2006), it is composed of entities (real or abstract, graphically presented as rectangles), their attributes (characteristics graphically presented as text) and their relationships (graphically presented as lines that connect the rectangles). The most significant difference between IDEF1 and IDEF1X is in their definition of the entity which, in IDEF1X, refers to a collection or set of similar data instances that can be individually distinguished from one another (IDEFX 2013). The model itself is presented by a graphical notation that represents the structure and semantics of information within a system (Kacprzak and Kaczmarczyk 2006), but without a possibility of modelling knowledge or fuzzy information. IDEF3 is used for the description of the process flow. Plaila and Carrie (1995) described it as the modelling method with the possibility of discerning what estimated system optimization is and what a prediction model of processes within the system is. It presents the behavioural aspects of an existing or proposed system (IDEF3, 2013). The most significant difference between IDEF0 and IDEF3 is that the latter does not let us create a model of the system, but captures precedence and causality relations between situations and events in a form that is natural to domain experts (Plaila and Carrie, 1995). IDEF3 models are built of four different groups of elements. First group elements define objects, time windows and existing relations. Elements from the second group form logical mechanisms that describe any existing relations 33

56 among the elements from the previous group. The third group consists of linking elements that create behavioural possibilities for the elements from the first group, and the fourth group presents elements that hold any additional information about the elements from the first group. Role activity diagrams (RAD) capture an entity s role presented within a business system. As stated by Badica (2003), the approach is a popular visual presentation of modelling focused on business processes. A single role includes a range of activities with the ability to launch one or more responsibilities of an entity. Murdoch and McDermid (2000) describe its graphical notation as a composition of grey, rounded boxes that represent the role and carry its name. Black boxes represent tasks that must be named and connected with vertical positioned lines (they represent the state of a role). In addition to these, modelling notation consists also of triangle symbols that represent current and alternative sequences of tasks within the scope of a role. White boxes represent interaction which is, in fact, communication between the roles. A RAD provides an excellent means of describing dependencies between roles in organizations that work discretely and in unison to achieve a goal (Cox et al., 2004) As argued by McCarthy (1982), Resources, Events, Agents (REA) is an approach that creates a process model consisting of an enterprise-wide part that represents the global view and a set of local views. Its graphical presentations consist of boxes that represent entities, diamonds that represent association relationships, and three-dimensional elements that represent generalization relationships. The intuition behind the core concept is that every business transaction can be seen as an event where exactly two agents exchange resources (Schuster and Motal 2009). O Leary (2004) identified its deficiency as being unable to model tasks that are normally included in the reality that the model represents. He also proposed a way of including tasks as event information that, according to his belief, would be simple to design and at the same time importantly expand modelling capabilities and add the ability for business process improvement and reengineering. The REA approach can be used for a topdown analysis as well as a bottom-up analysis (Geerts et al., 2001). A resource diagram that represents different levels of the organization creates a value chain out of the most important business processes. Kim and Kim (1997) define the dynamic process modelling method as an approach constructed of three components: performance-based process redesign guidelines, dynamic process model (based on the existing modelling notation), and performance evaluation visual 34

57 simulation (based on existing simulation environments). They find it useful especially when answering the demand of overall modelling or redesign. It delivers animated simulation, based on the model that enables the performance of evaluations. The graphic visualization plays a key role in communication related to business processes with employees or management members that are not IT experts. A comparison between different dynamic process modelling implementations, such as the dynamic model for the activated sludge process with high strength wastewaters (Costa et al., 2009), the dynamic model of the process of protein synthesis in eukaryotic cells (Skjøndal- Bar and Morris, 2007), and the dynamic model of the mood updating process in consumer behaviour (Hollbrook and Gardner, 2000), confirms the explanation provided by Joo et al. (2001), who highlighted the most important difference, comparing it with conventional process modelling in instructions which define how a dynamic model works. In dynamic modelling, these instructions have to be created in real-time by reflecting the dynamic of system reality. Tabular application development (TAD) is an object-oriented methodology that uses various tables to represent the functioning of a system; each illustrates a specialized view of an organization and its processes (Damij, 2007). According to Damij (2000), the TAD approach results in tabular description and presentation of the business process environment. It consists of an entity table that defines and describes all subjects and their associations within the business process. This table is named the activity table and maps every activity that operates as part of the process. Task table is designed to define the connection between entities and tasks within every activity in the process. In the table, activities are set vertically and entities horizontally, creating a matric system. When activities are allocated to selected entities, the table cells are filled with the chosen notation based on characters or symbols, creating activity flow. The Petri net (PN) can be described as a bipartite diagram with a defined direction and two types of objects: places and transitions. The nodes are connected via directed arcs. Connections between two nodes of the same type are not allowed. Places are represented by circles and transitions by rectangles (van der Aalst, 2000). When every place incident on a transition is marked, that transition is said to be enabled. An enabled transition may fire by putting one token into each of its output places (Hemmje et al., 2011). Bobbio (1990) describes Petri net as a graphical visualization of activity flow within a complex system, and 35

58 by comparing the approach with similar modelling techniques, found Petri net as an appropriate method of describing natural, logical interactions among parts or activities in a system. The typical situations that can be modelled by Petri net are synchronization, sequentiality, concurrency and conflict (Bobbio, 1990), and they have been applied mostly in manufacturing and safety-critical systems (Peleg et al., 2005). Flow charts are simple diagrams used for documenting algorithms or processes in a formal, graphical way. Process steps are displayed in boxes that are connected by directed arrows (Winkelmann and Weiß 2011), while Turner (2006) adds also pictures and engineering symbols to its notation. Rosen et al. (2009) describe it as a mapping tool that presents a sequences of steps and decisions needed to accomplish procedures that form a business process. It shows all of the inputs, such as subassemblies and materials, machining processes and machine operators basically anything that happens in the process and all of the outputs of that process, including the finished product or the individual manufacturing or handling steps along the production line (Lorenzi 2007). Turner (2006) finds flowcharts a basic but useful tool that is especially suitable for providing a common reference point when discussing and analysing a process, facilitating understanding of a common process or one that appears disordered, and supporting the understanding of relationships and time sequences within a process. At the same time, flow charts offer the ability to identify problems, bottlenecks, patterns and relations within the process. Van der Aalst (1999) and Dumas et al. (2005) describe the event driven process chain as an readily understandable visual representation of a business process with the aim to describe the process logic on the business level of the organization in a way that broadens the usability from business process experts to other business people. Visual representation is composed of the following groups of elements: functions (represented as rectangles, they are the basic building blocks and present activities within the process), events (represented as hexagons, they stand for situations before or after the function, are used to link functions, and are presented within the process as a point in time), logical connectors (represented by a circle, they are used to connect activities with event). In addition, the event driven chain must follow additional rules, such as: proper naming of events, clarification of the beginning and the end of process, it must contain at least one function, and can be composed of several chains. Each event-driven process must follow some simple design rules to avoid or reduce undesirable behaviour like deadlocks right from the beginning (Dumas et al., 2005). For this reason, there 36

59 is no strict or complex set of rules that would predefine a process design. Such approach provides users with a full scale design without any strict limitations. Gantt charts are widely used to represent production plans, schedules, and actual performance (Jones, 1988). Pankaja (2005) defines a Gantt chart as a graphical presentation of the work breakdown structure, combined with total time duration defined for a single task, resources delegated to the tasks, and overall process completion percentage. Gantt gave two principles for his charts: one, measure activities by the amount of time needed to complete them; two, use the space on the chart to represent the amount of the activity that should have been done in that time (Herrman 2010). Wilson (2003) further developed the definition, explaining that Gantt chart logic primarily uses systemic solutions and pays no attention to algorithmic solving of the problem. In a Gantt chart, each task takes up one row. Dates run along the top in increments of days, weeks, or months, depending on the total length of the project (Pankaja, 2005). The expected amount of time that will be needed for every single task to be carried out is presented by horizontal lines that, at their beginning and end, define also the start, duration and completion of the task. Business Process Modelling Notation (BPMN) is a modelling approach that consists of the graphical presentation model, describing the business process on the control-flow base with the core graphic elements and an extended specialized set (Ouyang, 2008). Recker (2010) defines the core elements as a set that provides a graphical presentation for substantial business processes in simple, intuitive models, while together with extended specialized set, they enable the presentation of advanced problems, such as orchestration and choreography, workflow specification, event-based decision making and exception handling (Recker, 2010). The approach is easy to understand and clear to interpret, however, according to Lerner et al. (2010), it is not supported by the formal semantic background that was proven to be risky as different process models are not understood equally, or designed almost identically, but with discrepancy in their core meaning. A similar shortcoming is described by Recker (2010), who mentions a lack of precisely defined and integrated common business rules, resulting in an additional deficit that the user is forced to improve by himself. Unified Modelling Language (UML) is a visual modelling language for modelling system requirements, describing designs, and depicting implementation details (Siau and Cao 2001). Siau and Cao (2001) further define UML as a member of the object-oriented methodology group that was developed and standardized to stabilize the modelling confusion among users. 37

60 Opdahl and Henderson-Sellers (2002) stress that the method is able to provide users with solutions to specification, visualization, construction, documentation and evaluation software dynamics, information flow and business processes. They describe it in more detail as a set of graphical tools for developing diagram notations and solutions that visualize a static model of process performance, usability and architecture. As defined in UML superstructure specifications (2013), the approach contains two categories of diagrams. The first represents structural information and the second can be used for presentation of behaviour and interactions. Erickson (2008) pointed out its negative sides as being difficult to comprehend due to semantic and notation problems, and as being complex and not user friendly, but concludes that in spite of this, UML has become a much used and useful tool for many organizations and people, and in some ways is now the lingua franca of modelling (Erickson, 2008). Web service business process execution language provided the basis for the development and execution of business processes that are distributed over the network and available via standard interfaces and protocols (Paci, 2008). The approach, described by Huang et al. (2009), is in fact a precisely defined specification language with syntax based on XML that defines the activity flow of the single process and interaction with other processes within the system. It is capable of defining fifteen activity types, including: process (composed of activities), activity (basic or structured), basic activity (receive, invoke, reply, assign, throw, terminate, compensate, wait, empty), and structured activity (sequence, switch, flow, while, pick, scope). Barros and Dumas (2006) find the reach of WS-BPEL as very localized, and criticize its limitations in covering the global picture of particular environments, because such gap makes it less applicable when a need exists for a global picture of a process occurring in a supply and distribution network, rather than multiple small pictures focusing on individual services or service roles (Barros and Dumas 2006). Kang et al. (2007) describe the web service choreography description language as a modelling language using XML syntax and having the ability to describe cooperation or any connection among multiple subjects, services, or processes in interaction with the aim of achieving a common goal. Instead of describing cooperation from the point of view of involved parties, WS-CDL approach rather takes a global or neutral perspective. This ability is what makes the approach suitable for the web service business process execution language. Mendling and Hafner (2008) distinguish two main concept parts of the model language. Package information is the core of any service choreography and covers definitions of information and 38

61 process correlations. Choreography, on the other hand, is defined a set of rules that define collaboration and other subject interactions. It can stand for a single choreography or multiple choreographies at the same time. A business process modelling language (BPML) is a standardized language-based code for describing a business process (Edgar, 2004). Edgar (2004) further explained that the BPML language, based on XML schema, provides the complex semantics and syntax for business process modelling and can enable business processes to be described and managed independently from the software used to implement and support them (Lee 2005). Languages like BPML provide a link between the typical process designer s flow charts, process maps, and executable computer code (Lee 2005). Van der Aalst et al. (2002) identified the following main structural components of BPML: activities (simple and complex, they stand for specific functions), processes (built by several different activities, predefined by process hierarchy: top-level process, nested process, exception process, and compensation process), contexts (they define a scenario in which activities get executed), properties (their main task is information exchange, they can operate only within the context), and signals (used for coordinating the execution of activities). Hofreiter et al. (2002) described the most important feature of ebxml as the ability of creating a global e-market where different parties can communicate, trade and create business deals, while Kim (2002) argued that ebxml enables all involved parties to get the information necessary to present a business process, without a complex analysis or modelling. Summed up from Patil and Newcomer (2003), ebxml architecture consists of five main elements that enable the modelling of the following business scenarios: business processes, messages, agreements, business registry and the core component that supports a user in establishing the necessary documentation about the business process model. Business process modelling is the groundbase of any sophisticated business process management approach. The process model gives visual assessment of the processes and at the same time already indicates possible optimization locations or patterns within the process architecture. 39

62 2.2.4 Business process simulations Simulations as known today became, according to Paul et al. (1999), an environment that offers a powerful mechanism for dynamic business process modelling aiming to conduct either improvement or innovation. Doomun and Jungmun (2008) define both improvement approaches as to-be processes, meaning that they are put into the future always with some level of uncertainty, making simulation a very beneficial tool for predicting and evaluating the planned optimization results. Modelling, analysis, simulation and improvement of processes are on the increase as only a thorough comprehension of the processes within on organization can lead to effective, efficient and value-adding systems. Furthermore, conceptual modelling of processes is deployed on a large scale to facilitate the development of software supporting the processes, and to permit their analysis, re-engineering or improvement (Chen et al., 2012). Processes are modelled with the aim of analysing their current states within the organization, as well as of improving them through the execution of potential what-if simulation scenarios. Currently, few organizations maintain a formal model of their process network (Damij, 2000). Even so, these formal models were usually the product of a series of interviews, reviews, examinations, etc. conducted at a time when the organization encountered apparent or serious problems. Construction of a process model makes use of several accepted modelling constructs: delays associated with performing work, statistical distribution of these delays to represent the observed variability, dependence of processes on completion of earlier processes, queuing of input entities waiting to be processed, decision logic that directs entities into alternate flow paths depending on their characteristics, application of resources to the work of a process, and the costs associated with these resources (Damij, 2000). Consequently, an as-is model encompasses the above constructs with the aim of imitating the real process under inspection within the organization. As claimed by Damij (2006), simulation modelling is based on very simple principles: the analyst builds a model of the system of interest, writes a computer program which embodies the model, and uses a computer to initiate the system s behaviour when subject to a variety of operating policies. Furthermore, it is thought that simulation modelling extends the potential of process modelling and analysis. According to Damij (2000), modelling and simulation of a process network serve three immediate purposes: organizing the results of interviews and research, identifying the cause 40

63 of observed performance issues, and exploring alternate process network configurations that improve performance. The aim is to efficiently image the process network. Hierarchical modelling tools are the most useful, as they contain features like simultaneous understanding of high-level and detailed views, as well as the manageability of the model (Damij, 2000). Simulation is the imitation of the operation of a real-world process or system over time (Damij, 2007). A simulation model enables the analyst to observe and study the system s behaviour as it advances through time. Doomun and Jungum (2008) claim that simulation represents a powerful approach for analysis and quantitative evaluation of processes. They classify simulation models into three groups, depending on their attributes. - Static or dynamic. A static model is a model where time within the real process is insignificant, while a dynamic model incorporates changes over a period of time. - Deterministic or stochastic. A deterministic model is defined by a sequence of events such as input recognition enabling the output definition, whereas a stochastic simulation model has, according to (Damij, 2007), one or more random variables as inputs. - Discrete or continuous. A discrete model is a model that consists of discrete events which are events that happen at particular times, and a continuous one consists of variables changing continuously over time. Consequently, Doomun and Jungum (2008) state that processes are, in general, represented as computer-based dynamic, stochastic, and discrete simulation models which are defined as abstractions of the actual processes, represented in the computer as a network of connected activities and buffers through which jobs or customers flow, and must also capture the resources and various inputs needed to perform the activities. Discrete-event simulation describes how a system with discrete flow units or jobs evolves over time (Doomun and Jungum, 2008). Therefore, according to Damij (2007), discrete-event simulation examines the modelling of systems in which the state variable changes only at a discrete set of points in time. What differentiates a discrete-event model from a continuous one is the fact that it deals with the attribute time only when the event actually happens. Consequently, as explained by Doomun and Jungum (2008), such a perspective of events and time enables significant time compression because it enables skipping through all time segments between events when the state of the system remains unchanged. Simulation packages enable simulation runs of vast and various numbers of events that may in reality happen over a long period of time. A 41

64 discrete-event simulation model focuses on the state of the process at specific time points when the events occur. Hence, when executing the simulation run, the simulation clock jumps between the events and regards the system as staying the same in the meanwhile. A simulation model is normally based on a set of assumptions regarding the system s operation. These assumptions are expressed in mathematical, logical, and symbolic relationships between the entities, or objects of interest, of the system (Damij, 2007). After a simulation model has gained form and been validated, it is deployed in order to examine various what-if questions regarding the real-world system, so that any future alterations of the system are first simulated. As a result, it provides forecasts about the impact of the alterations on the system effectiveness. Researchers (Doomun and Jungum, 2008) summarized some of the main attributes that make a simulation powerful. Simulation, like analytical modelling, provides a quantitative measure of performance. However, unlike analytical and symbolic models, it is able to consider any kind of complex system variation and statistical interdependencies. Simulation is capable of uncovering inefficiencies that usually go undetected until the system is in operation. The availability of special-purpose simulation languages, massive computing capabilities at a decreasing cost per operation, and advances in simulation methodologies, have made simulation one of the most widely used and accepted tools in operations research and system analysis (Damij, 2007). Business process simulation is a necessity in the managerial approach of the modern organization. It helps understand the living dynamic of process architecture of the organization, identifying important process patterns and those parts of the processes that hinder an organization s operational effectiveness. Consequently, business process improvement measures can be applied based on using business process simulation. Such an approach makes it possible to verify the improvement-oriented solution before the real-life application. 2.3 Knowledge management The following definitions of data, information, knowledge, and wisdom attempt to capture the common essence of the various definitions presented in the knowledge management literature. Data are considered to be unprocessed raw representations of reality. Information is 42

65 considered to be data that has been processed in some meaningful way. Knowledge is considered to be information that has been processed in some meaningful way. Wisdom is considered to be knowledge that has been processed in some meaningful way (Faucher et al., 2008). It has been shown that data, information, knowledge, and wisdom could all be tacit or explicit, and that understanding is the basis of conversion processes among them. The classical knowledge hierarchy had to be expanded. Indeed, in order to attain the full scope of the knowledge hierarchy, it was necessary to add two concepts: existence and enlightenment (Faucher et al., 2008, p. 13). To make knowledge efficiently translated into value or profit for a company and strategically used by varieties of users, relevant knowledge management is self-explanatory (Gao et al., 2002). Knowledge management (KM) involves activities related to the capture, use and sharing of knowledge by an organization. It involves the management both of external linkages and of knowledge flows within the enterprise, including methods and procedures for seeking external knowledge and for establishing closer relationships with other enterprises (suppliers, competitors), customers or research institutions. In addition to practices for gaining new knowledge, knowledge management involves methods for sharing and using knowledge, including an establishment of value systems for sharing knowledge and practices for codifying routines (Ringel-Bickermaier, 2010). Knowledge is recognized as one of the drivers in technological and development-oriented organizations and can be found in different forms in any organizational system, even though knowledge management still lacks efficient implementation in environments where knowledge is being overtaken by the work-oriented approach. Therefore, it is important to assess different organizational systems from the knowledge perspective, to test various methodologies in unconventional environments and to develop new, tailored methods Organizational knowledge Skovvang, Christensen and Bang (2003) see the creation and dissemination of knowledge as a continuous process between people and technology, having a solid position in a processoriented organization. KM programs succeed when knowledge capital is employed to accomplish specific business strategies. There is no successful KM program which is not tied to a specific business strategy or goal (Coopers, 1999, cited in Plessis, 2007). Business organizations have an inherent interest in using both the business knowledge owned by the 43

66 organization, and personal knowledge of their employees (Gao et al., 2008). In order to accomplish this, firms must develop an absorptive capacity the ability to use prior knowledge to recognize the value of new information, assimilate it and apply it to create new knowledge and capabilities (Gold et al., 2001). KM implementation requires changes in organizational philosophy as it forces an organization to redefine its beliefs system, conventional workflow, power structures and technology utilization (Bhatt, 2001; Glazer, 1998; McKenzie et al., 2001). In conjunction with that indication, the implementation of knowledge management processes could alter the existing systems and procedures that apply in the organization (Rusly et al., 2011). A good example of such alteration can be found in external knowledge capturing such as crowdsourcing or customer relation system. There is no doubt about the fact that organizations must manage knowledge if they are to be successful, or even survive, in the marketplace. While this remains an accepted fact, one thing has changed over the last few years: the role of an organization in how knowledge is managed. This transformation is especially visible when it comes to managing knowledge from external sources. The most important sources are the customers (users), both current and future (Moon and Desouza, 2010). Knowledge management is thus a natural solution for improving operations and enhancing customer service (Seba and Rowley, 2010). Not only local, semi-closed organizational systems are exposed to changes that originate from knowledge management. Knowledge has always been globally present but locally concentrated. Through the global rise of knowledge importance awareness, we might expect also a shift in the existing distribution. We are undergoing a global shift towards a knowledge-focused economy. For example, Walczak (2005) claims that the worldwide economy has shifted from an industrial manufacturing/product orientated economy to one based on knowledge and services, where the principle commodity is information or knowledge. Due to a new borderless society, the importance of knowledge management is being realized and identified as the critical success factor for today s businesses (Lim et al., 1999, cited in Wadell and Stewart, 2008). Therefore, the organizational knowledge has an important role in the lifetime of the organizational system. It is an influence that has the capacity to influence organizational change. To be able to predict, monitor and control change, the aim of my research was to 44

67 develop a knowledge management approach, feasible in the chosen organizational system and providing the capacity to map and measure organizational knowledge Tacit, explicit and experiential knowledge in the loosely coupled system Knowledge is an important factor which influences dynamics of the workflow within the assessed flood response loosely coupled system. In order to determine how knowledge and learning integrate into the system, their theoretical definition must first be explained. Theoretical research discusses two general types of learning: implicit and explicit (Jasimuddin et al., 2005), distinguished by Kirkhart (2001) according to the learning intention. The implicit learning process is based on unintentional learning, while explicit learning process must include an intentional effort to attend learning. Pozzali (2008) links implicit learning with the emergence of tacit knowledge, first described by Polanyi (1958) as all the knowledge that cannot be codified in explicit form, such as formulas, figures, manuscripts, etc. Yang and Li (2012) associate explicit learning with recalling memorized rules and codified knowledge, while Sun (1994) presents it as the process of precise reasoning directed by a set of rules. Such a learning process yields knowledge which can be articulated, written, codified, evaluated and described (Helie and Sun, 2010). Experiential learning, also detected in the assessed flood response loosely coupled system, as described by Kolb (1984), is a process where knowledge emerges through capturing and reshaping an experience. Mainemelis et al. (2002) expanded the previous definition with a model which identifies two dialectically connected ways of capturing an experience. The first is a concrete experience and the second is abstract conceptualization. In addition, the model provides two dialectically connected ways of experiencing change: reflective observation and active experimenting. Based on the model, different styles of experiential learning can be identified that favour the desired ways of capturing and interchanging the experience among the listed dialectics. A question which arises is how long does experiential learning remain of experiential nature. Illeris (2007) provides a definition, claiming that experiential learning is a process where learning dimensions, incentives and interactions are significantly subjectively balanced. However, it is still not clear when experiential learning tilts more towards experiential nature than towards traditional learning, based on the teaching or reading processes. The author 45

68 provides an answer, declaring that the more the subject of learning is complex, the higher the possibility that the learning process will be experiential. Burnard (2005) delivered a comparison from the perspective of experiential knowledge, which emerges because of experiential learning. He argued that experiential knowledge is the opposite of a priori knowledge, which, as claimed by Giaquinto (2008), emerges before the experience and is consequently independent of the experience. Therefore, I am dealing with posteriori knowledge, which is, as presented by Müller-Merbach (2007), divided in five groups (Figure 2.1): - knowledge about man-made structures (laws, constructions, forms, music, traffic system, etc.), - knowledge based on scientific exploration, - knowledge composed of empirical facts (the average distance between the Earth and the Sun is 149 million km), - knowledge based on social discoveries (social science laws), - knowledge about behavior assumptions (assumptions about people, their wishes, goals, needs, etc.) Figure 2.1: Five groups of posteriori knowledge Source: Müller-Merbach (2007) 46

69 Experiential knowledge, detected in the flood response loosely coupled system assessed in the research, refers to all three levels described in Figure 2.1, but only on the right side of the graph, which is, as described by Müller-Merbach (2007), more unclear and conditional compared to the left side Experiential community learning Experiential learning of communities is the main source of customer knowledge within the flood response system. Therefore, it represents an important complement to the conventional learning channels and knowledge sources. Experiential community learning during and after flood events enhances disaster preparedness not only of private households affected by floods, but businesses and authorities as well (Kreibich et al., 2011). Experiential learning is a learning process where experience plays a central role (Kolb, 2000). For such a process, it is of great importance that it be implemented in the community of interconnected partnerships (Benecke, 2011) because community-based learning initiatives, which are experiential and action-oriented, complement the regular forms of learning (Jakubowski and Burman, 2004). They provide learners with the chance to participate in organized activities and to meet the needs of the community (Tapps et al., 2014). Experiential learning is a very effective approach to disaster threatened learning communities (Rijumul et al., 2010). Well-managed and applied experiential community knowledge can play a vital role through ensuring the availability and accessibility of accurate and reliable disaster risk information when required (Pathirage et al., 2012). In general, people tend to ignore personal disaster risks, seeing themselves as immune to disasters (Slovic et al., 1981). Such optimism is present in communities with a low disaster risk where people lack direct experience with disasters (Weinstein, 1989). Experiential learning thus improves risk perception in general and flood risk perception in particular (Bosschart et al., 2012; Grothmann and Reusswig, 2006; Siegrist and Gutscher, 2006; Terpstra and Lindell, 2011). Experiential community learning which occurs as single-loop learning, can result in changes of community behaviour, strategies, and techniques (Hayward, 2007). Alternatively, it can enhance partnerships among communities and public services responsible for disaster preparedness and response (Gamboa-Maldonado et al., 2012). Experiential learning enables a 47

70 community to verify its expectations of the governing bodies capabilities to implement an efficient disaster response process through its ability to influence the response process (Chamlee-Wright, 2009). Such collaboration during disaster events is of high importance for a community in terms of developing its resilience and withstanding the burden of the ongoing emergency (Plough, 2013). For this reason, community reactions have a crucial role in process optimization during disaster events (Shughart, 2011). Experiential community knowledge can be compared to customer knowledge when a community assumes the role of an entity demanding a service or a product. A community learns through experience gained from service or product use, just like a customer does when using commercial services or products. Therefore, experiential community knowledge surpasses the status of just a rounded knowledge hub of an informal organizational system, becoming customer knowledge with the ability to trigger influences outside the organizational system Loosely coupled process architecture and organizational learning Process architecture presents a structural description of the processes within an organizational system, including all process components, such as: inputs, outputs, activities, entities, events, links and any other specific element of the organizational system. Ionita (2011) describes modern successful business organizations as being strongly dependent on their business architecture in order to conduct their daily activities. Such enterprises boast a stable organizational system that is also capable of intended and controlled flexible adjustments. The process architecture supports both aspects of the system and must be robust enough to maintain the delivery of desired process outputs. Barros and Julio (2011) argue that the design of relations among process components is one of the most important factors of process architecture. Recognizing organizational knowledge as one of the pillars of the organizational system, process architecture design is consequently not able to provide sufficient process outputs without proper knowledge management. Niu (2010) highlights the relationship between knowledge management and organizational adaptation, reflecting the relation between organizational knowledge and process architecture and indicating an influential connection between them in the unstable organizational system. An important characteristic of the loosely coupled systems process architecture is constant change due to the dynamic interaction between activities and weak links between entities, 48

71 which creates several different process models on a timeline where a usual business enterprise would create its outputs without a single model. Such loosely coupled process architecture evolves with a high tendency to achieve output priorities of the organizational systems, consequently including and discarding different activities, which reflects on a wide range of emerging activity flows. According to Martinez-Leon and Martinez-Garcia (2011), the less formal and less centralized organizational systems enhance the organizational learning process. Conversely, an open, less rigid, loosely coupled organizational system creates an open environment that encourages organizational learning, but also creates a need to assess more complex, less transparent and harder to follow learning processes. Organizational learning in a loosely coupled organizational system adapts to its process architecture and changes in activity flows in a reactionary and interventionary manner that is, according to Tennant and Fernie (2013), similar to an underdeveloped knowledge management application within industrial enterprises. Rapid change of activity flows constantly includes and bypasses activities and procedures that connect or break away from process architecture, not creating only new variants of operational business processes, but, as argued by Firestone and McElroy (2004), creating also knowledge processes and processes for managing knowledge processes. Knowledge management process follows an organizational learning process, influenced by changes in activity flow. According to Popper and Lipshitz (2000), such interaction is an important survival characteristic of an organization in dynamic environments. Loosely coupled systems use internal and external knowledge to run their processes while creating new knowledge that emerges based on activities included in their process architecture. Although such a concept is similar to the description of a learning company proposed by Pedler et al. (1989), who described it as an organization that constantly changes in order to achieve its goals, Garvin (2000) suggests that a learning organization does not lose critical knowledge when key knowledge sources leave it. A loosely coupled system and its organizational learning strongly depend on the state of its process architecture that is much less stable than a usual learning organization and its learning process gets more affected by changes in its operational processes. Patterns that occur within process architecture of a loosely coupled system are the key knowledge source for organizational learning in a loosely coupled system. Process patterns are important because they influence organizational learning in several different ways. Barros (2007) finds process patterns as architecture widgets that include activities, activity flow that connects them, and the business logic that determines the operative run of the process. They 49

72 are subjected to constant change, but when they settle in the state of operation, they appear to be the important learning spot. Patterns that arise within the architecture and create bottlenecks or otherwise negatively influence the process architecture have a tendency to collapse, the system excludes them from process architecture to maintain a sufficient level of operability. However, knowledge also emerges as the basis of information about a particular process pattern and the circumstances that contributed to insufficient process behaviour. Russell, van der Aalst, and ter Hofstede (2006), cited in Lerner et al. (2010), categorize patterns into four domains: control flow, data flow, resources, and exception handling. Considering these four domains in light of the definition set by Barros (2007), it becomes clear that patterns of every domain include activities, activity flows and business logic, but create different intermediate outputs according to their domain. Consequently, the knowledge that emerges in this pattern is created through the same learning process, but in a different domain of the process. Knowledge management process should consider different domains of emerged knowledge through the same principle. It is important to distinguish among different knowledge domains to be able to evaluate the knowledge and further expand it. It is also important to have the ability to record and contain knowledge, as well as to trace and recognize where it originates and at what point of the operation process it can be implemented. Due to highly unstable process architecture, it is very important to maintain a knowledge management solution with the ability to react to any process pattern occurring within the architecture, and to assess the consequences of its possible exclusion. Only a responsive and dynamic knowledge management solution ensures the possibility to successfully recognize, monitor and control organizational learning in a loosely coupled system. The knowledge management solution described above creates the ability to recognize and evaluate different knowledge sources that originally appear as entities in the process architecture of a loosely coupled system, but also possess the knowledge needed to run the operation process or the ability to navigate the learning process in the system. Such a managing approach is inevitable when we want to understand the learning dynamic of a loosely coupled system and the characteristics of knowledge which emerges within its process architecture. 50

73 2.3.5 Open knowledge management approach to a loosely coupled system A learning process in a loosely coupled system is subjected to activity and process patternbased learning that relates to the operating part of process architecture. Moreover, organizational learning also strongly depends on the entities included in the process architecture. Being highly unstable, a loosely coupled system frequently changes the set of its entities, mostly due to pattern-based learning and consequent recognition of efficiency gaps. In order to include all possibly important entities that appear as a knowledge source within the system, the knowledge management solution must have the possibility of allowing their integration even though some entities might appear in an unexpected form. This is why openness provides very important characteristics of the knowledge management system in a loosely coupled system. Awazu and Desouza (2004), cited by Lee and Ge (2010), connect open knowledge management with learner s own initiative, participation in the process of creation, organization, assessment and sharing of the relevant knowledge. Organizational learning in a loosely coupled system approaches the characteristics of an open management system in organizations with stable processor architecture, although its openness must be considered also with the ability to apply the listed features to a wider range of entities, knowledge inputs and consequently types of knowledge, information and data, to ensure the openness and to contribute to the creation of more and better knowledge (Garcia- Penalvo et al., 2010). Frequent changes of the loosely coupled system and possibly also process patterns result not only in learning about the operating process, but also in learning about the knowledge that emerges throughout the process. Connell (1995) describes the output of such a learning process as metaknowledge, i.e. knowledge that is fluid and changes with the learning of its source entity. Organizational systems with clearly expressed loose characteristics inevitably meet the repeating process patterns that create metaknowledge based on several loops of the same activity pattern within the process architecture. Freeman and Knight (2011) find that such learning process and knowledge are created as an output, as an expansion of the base knowledge, higher awareness and discipline that is, in a highly unstable organization, very important for its own sustainable existence. Together with internal or external knowledge, a loosely coupled system constantly faces also several information flows which consist of different types of information that contribute to the 51

74 operating process of the system. In the same manner as knowledge can be reused (Majchrzak et at., 2004), through the metaknowledge within the organizational system, information and metainformation must be considered as an important part of the knowledge management solution. Maier et al. (2011) understand metainformation about process inputs and operating methods as an important organizational widget that eventually merges into the pool of knowledge within the system. Until the transformation of information into knowledge does not take place through the implementation in the operating process or internal process patternbased learning, there is no tangible metainformation or metaknowledge about its usefulness and importance. Due to unpredictable change of the loosely coupled system, such information must be easily contained, and must be ready to be engaged in the operation process at any time when needed. Burns et al. (2010) distinguish different entities within the organizational system, based on the level of the knowledge they contribute to the system. According to them, the most knowledgeable entity of the system is the architect, who can recognize and identify different process inputs and outputs and predict influences on the process output when input variable changes. A loosely coupled system faces more dynamic entity turnover, which means not only that the entity which has the ability to become a system s architect might not be included into the system, but that it could even be excluded due to a collapse of process architecture. Consequently, the system is faced with a possibility to precede into the state when its processes would operate without a general understanding where the knowledge that triggers the inputs originates. In case of high fluctuation within the organizational system and the possibility that no entity will have the capabilities to take over the role of the system s architect, a knowledge management solution must provide an efficient compensatory mechanism with the ability to track, identify, assess and monitor knowledge inputs influencing process architecture of the loosely coupled system. Jiang et al. (2010) summarize knowledge inputs as research and development capital, human resources, while Whitely-Clarke and Teare (2011) describe knowledge inputs as action-based learning, that, in the loosely coupled organizational system, leads to the possibility that not just knowledge patterns, but every single activity within the process architecture holds the potential to trigger action-based learning and become a knowledge input for the system. 52

75 Complexity of knowledge management from a knowledge input perspective in a loosely coupled system grows much faster that in a stable organization. Ahrweiler et al. (2011) find such system to be knowledge-intense with intervening factors, self-organizing interaction patterns, strategies of the individual entities and different, innovative organizational frameworks. A loosely coupled system has the ability to operate in knowledge-intensive and stable organizations, so, in fact, it is not necessary that a loosely coupled system would evolve into a knowledge intense system nor that it would be capable of keeping knowledge-intense characteristics. Due to such uncertainty concerning organizational knowledge and the learning process, a knowledge management solution must be capable of operating at different stages of organizational learning in order to achieve knowledge management maturity that can be, according to Kruger and Johnson (2011), related to organizational performance. Considering the unavoidable characteristics of the loosely coupled system, summarized as low process maturity, the knowledge management solution must provide the opposite performance to be able to operate within an immature environment and to deliver the desired results. Arling and Chun (2010) argued that a knowledge management solution reaches an appropriate maturity level when progressing through different maturity stages together with the organizational environment. A loosely coupled system has no ability to operate as a mature organizational system because it loses its primary characteristics as such, but at the same time a knowledge management solution that controls knowledge sources, inputs and learning processes, creates a stake not to change the looseness of the system or improve its process maturity, but to ensure it operation with proper knowledge support in order to achieve the necessary outputs. The implementation of such a solution, which is capable of remaining at the appropriate maturity level and deliberately adapting to the state of the surrounding organizational system, could define a loosely coupled system as a knowledge-driven environment Knowledge mapping The idea that knowledge needs to be codified is central to many claims that knowledge can be managed (Hall, 2006). Knowledge codification particularly the emergence and use of codes and the ability to decodify them provides a theoretical basis for explaining what it is which both enables and limits the communication of knowledge (Hall, 2006). Knowledge 53

76 codification, together with diffusion, is emphasized also by Lytras and Pouloudi (2006), who take Boisot s (1987) model to describe the importance of diffusion and codification as characteristics of knowledge. Knowledge maps have been recognized and used as an appropriate methodology to understand the relationships and interactions between knowledge stores and dynamics in projects and organizations (Yun et al., 2011). Driessen et al. (2007) find knowledge mapping important for higher transparency of knowledge available within the organizational system that provides insight into its qualities (Nyberg et al., 2003, cited in Driessen et al., 2007) and supports the problem-solving process. Davenport and Prusak (2000) define a knowledge map as a collection of data that does not contain knowledge but points to it. Knowledge mapping is a consciously designed communication medium, using graphical presentation of text, stories, models, numbers or abstract symbols between map makers and map users. Knowledge maps are excellent ways to capture and share explicit knowledge in organizational context, similarly to a geographer s map (Wexler, 2001). The importance of knowledge mapping is highlighted also because some knowledge cannot be captured or transformed in sustainable form. The nature of any business, the degree to which it depends on technology to perform its business, the extent to which it is willing to provide information mapping to its working community, and the methodology for maintaining documentation or database integrity in the future, must be defined in detail. In addition, there is the most unpredictable element of all, the human factor. It is impossible to capture on paper or electronically the tacit (instinctive, intuitive) knowledge of an individual, and the centrepiece of an enterprise is the know-how of organization s workers rather than the tools of production (Guy, 2004). In this case, knowledge mapping provides information about knowledge and the ability to assess its life cycle within an organizational system. According to Nonaka et al. (2000), knowledge assets, namely the inputs, outputs, and moderating factors of the knowledge-creating process, are extremely critical to the knowledge-creating processes. For proper understanding of how knowledge assets are created, acquired, and exploited, Nonaka et al. (2000) categorize them into four types: experiential knowledge assets, conceptual knowledge assets, systemic knowledge assets, and routine knowledge assets (Rai, 2011). Wang et al. (2012) approach the assessment with a knowledge management audit process, arguing that it not only has to evaluate knowledge demand, knowledge flow, existing knowledge property or resources, and future demand of 54

77 knowledge, but should also evaluate organization strategy, leadership, cooperation, study culture, and technology foundation (Wang et al., 2012) Discrete dynamic knowledge mapping The capability to maintain a stable knowledge management approach within a loosely coupled system as a very unstable organizational environment, comes from being able to detect, follow and assess the changes of the system s architecture that affect the knowledge aspect of the system or have been caused by the knowledge dynamics itself. Hall (2006) describes such a method as knowledge codification and finds it as a backbone of the ability to manage the organizational knowledge. Being able to codify the knowledge leads to easier diffusion, but according to Mahroeian and Forozia (2012), different types of knowledge demand different approaches. Taking into account the rapid changes of knowledge sources, inputs and different types of entity tailored learning, the knowledge management solution for a loosely coupled system must possess an intuitive codification mechanism to be able to successfully manage the knowledge within the system. Knowledge map is the solution for this dilemma and is, according to Driessen et al. (2007), of high importance for transparency of knowledge within the organization while acting as a solid problem solving support. Wexler (2001) identifies components of a knowledge map as: graphical presentations of textual content, models, numbers, and symbols that sufficiently cover organizational needs to codify its knowledge. When implementing knowledge mapping into the loosely coupled system, the design must be set in order to be able to process and map any knowledge-based or learning-based change within the organization. Replacements in the list of entities, adjustments in process architecture, and influences from external knowledge inputs will challenge the mapping concept constantly. It is important that all changes be successfully detected, assessed and included on the map in the standard form that must be flexible enough to proceed with any kind of systems modification. Kang et al. (2003) base the knowledge map concept on the workflow of a process architecture. Such solution partially corresponds to the needs of knowledge management solutions for loosely coupled systems with additional substantive upgrades that would connect workflow activities with other elements of the system. The ability to detect and track any organizational change that affects knowledge within the system delivers the possibility to maintain a constant view of the knowledge network 55

78 evolution process. Wang (2013) argues that even the influence of an individual within the system can impact the knowledge network evolution, while Taplin (2011) finds the learning network itself as a facilitator of organizational evolution. Thus, when we want to observe the evolution process of the knowledge network as well as its influence on the evolution of the loosely coupled system, dynamic knowledge map is a tool that would provide us with the ability to capture different evolution stages. Zhang et al. (2012) assert knowledge network mapping as a method used to recognize the evolution of topics that serve as knowledge centroids, which conceptually approaches the methodological need for mapping knowledge evolution in loosely coupled systems. In fact, in order to successfully implement a knowledge mapping methodology that will capture a knowledge network and its interference with an organizational network on the timeline of organizational change of the system, mapping process has to be discretized according to the ongoing changes within the system. With predefined rules that determine the borders for dynamic knowledge mapping intervals, it is possible to create several knowledge map prints with the same methodology, but at different times. Each print must consist not only of a graphical presentation of the map for intuitive and quick understanding, but also a numerical background in order to be able to quantitatively analyse the evolution process. The numerical component of the map must reflect every detail of the organizational structure and at the same time design the knowledge network of the system. Thus, I suggest the implementation of a relational database concept, used by Folks et al. (2012), and its upgrade with a semantic query, developed by Sedighi and Javiddan (2012). Not only can such a concept easily track and record any desired detail of the system and its knowledge network in the required time intervals, it also provides simple query requests to obtain the data for specific analytical purposes. According to Ahmed and Karypis (2012), dynamic networks have been recognized as an important knowledge source, providing us with the significant insights regarding patterns within the organizational system. The ability to conduct several different statistical methods of the overall database provides a more accurate overview of the knowledge network evolution and patterns that influence other structures of the loosely coupled system. What is more, this solution integrates the ability to convert the databases into ontology that, with its graphical representation, complements the insights of quantitative analysis. Such a relation between the numerical and semantic components enables smooth dynamical interaction that provides a constant record analysis of 56

79 the ongoing evolution. Beside the as-is analysis, dynamic knowledge mapping can be used to create a to-be ontology based simulation model, similar to the modelling solution of Kwon et al. (2010). With the possibility of predicting knowledge network patterns, the analytical ability is expanded not only to evaluate past or current states of the organizational system, but also to determine which knowledge network patterns should be encouraged to improve work processes in the system. In order to develop such dynamic knowledge mapping based on several interconnected components and to integrate the solution in the entire knowledge management concept of the loosely coupled system, the smoothest solution is to develop an open software solution. Such software will be flexible and robust enough to process constant change and provide the required quantitative analytical solutions as well as an intuitive graphic interface. With an extensive theoretical review, I was able to determine a gap in the application of business process management methodologies and knowledge management methodologies in loosely coupled systems. Even though both fields have produced numerous research findings, extensively covering the details in the practical application on conventional organizational cases, transfer of these insights into a loosely coupled system has not been executed to a full extent. Several gaps exist in the understanding of how to capture the structure of loosely coupled systems and analyse their process architecture. Further, it us unclear how to measure and evaluate learning processes in the loosely coupled system, how to detect the influences between learning processes and business processes, and, finally, how to design predictions related to knowledge and process influences. Therefore, I need to look for answers by applying and adjusting conventional business process management and knowledge management methodologies to the loosely coupled system. 3 EMPIRICAL RESEARCH The empirical research section first introduces a selection of insights which serve as a bridge from the theoretical part to the empirical research part, focused on the flood events in the Lower Sava Valley and the additional missing person investigation case. Next comes the background information related to Hypothesis 1, Hypothesis 2, Hypothesis 3, and Hypothesis 4. I then explain the data regarding the mentioned hypotheses and, in a subsection, present the 57

80 methodology used to address every single hypothesis and the obtained results. Finally, background information for the fifth hypothesis is presented along with and data, methodology and results related to it. 3.1 From theory to practical applications There is no doubt that global climate change has increased the frequency of extreme precipitation events and may cause many more floods in the future. Losses of human life and property are recorded through flooding fields, carrying away buildings, spreading infectious diseases, etc. (Qi et al., 2014). Flood risk is expected to increase in many regions of the world (Cammerer et al., 2012), which indicates heavier exposure to consequences caused by flooding. Repeated exposure to natural disasters, such as floods, renders people more resilient and facilitates social connectedness that enhances a sense of place (Boon, 2013). A logical result of threat acceptance and rising resilience in the urban, suburban, and rural communities in flood-endangered areas is the development of risk management, capable of flexible adaptation and improvement (Zevenbergen et al., 2013). Even though disaster risk management framework is under the jurisdiction of national and local governments from the financial (Chang et al., 2013), legal (Nishikawa, 2010), and implementational (Kakumoto et al., 2011) perspectives, suburban and rural communities tend to rely on their own experience and information sharing (Ryan, 2013). These communities are more likely to assess an ongoing situation and build their response based on their own social network, which includes trustworthy information sources (Hagar, 2010). In addition to such community networking, suburban and rural communities show sensitivity to the geographical origin of the external information concerning the ongoing flood situation. Communities are more receptive to the external information with local origin than information from afar (Cohen, 2007). Efficient disaster risk management is a shared goal of the community, national and local governments, and disaster responding services (Metcalfe et al., 2012; Reaves et al., 2014). It is also a subject of active community participation in the disaster management cycle (Jahangiri et al., 2011). To achieve a solid and sustainable level of community-based safety, a community should encourage its individual members to participate in disaster management processes in order to develop knowledge-based prevention capabilities and risk reduction (Zhang et al., 2013). An active community has strong learning abilities. Therefore, it can 58

81 easily assess its needs, share information, and provide the necessary public influence (Rolland et al., 2010). A learning community is capable of overtaking traditional organizational structures and mechanisms when addressing disaster and risk management (Said et al., 2011) through a diverse range of organizational and professional resources that can be called upon to assist recovery (Johnston et al., 2012). There is growing recognition that flood risk exists not only in unprotected floodplain areas but also in areas behind levees (Huthoff et al., 2015). Areas that used to be flood safe have suddenly become endangered by unexpected flooding. Handmer (2003) argues that urban communities are vulnerable and resilient at the same time. He claims that the element of uncertainty will always be inevitably present. Similarly, Becker et al. (2014) established a connection between community resilience and risk perception. They found that different risk perceptions might cause different preparedness stands and natural hazard responses. Klijn et al. (2008) describe a switch in common European flood risk management from flood protection towards flood risk management. In addition, flood event prediction is of high importance for a sufficient implementation of flood risk management measures. Simulation models have become a widely used tool for risk assessment and disaster prevention. Rabelo et al. (2006) find simulation modelling useful as a decision-support environment for space range safety. Chen et al. (2014) used simulation to create typhoon compound disaster simulation. Ertem and Buyurgan (2011) created a simulation model to address the inefficiency problems in the procurement operations in disaster relief logistics. And last but not least, Garcia-Pintado et al. (2014) used simulation for satellite-supported flood forecasting. Within the general field of disaster modelling and simulation, flood prediction and computer model based assessment is a recognized methodological approach. Ahmad et al. (2000) support the idea that there is a strong need to explore realistic flood simulation techniques that represent complex dynamic systems. Jones et al. (2012) used computational modelling to determine efficient coastal flooding protection, while Kozhevnikova et al. (2012) found mixed probability distribution a useful tool to model fluctuations of the Caspian Sea. Dawson et al. (2005) successfully used computational modelling methodology to provide an efficient means of assessing the flood risk of a complex dike system. One of the useful approaches to flood simulation and modelling is the introduction of machine learning techniques (Srivastava et al., 2013), also labelled as artificial intelligence techniques 59

82 in the field of flood risk management (Sayers et al., 2014). Chang-Chi et al. (2008) consider decision tree algorithms to be a suitable tool for typhoon warnings. Practicality and usefulness of the data mining approach was presented also by Zhang et al. (2013), who used the C4.5 algorithm for the data mining approach to analyse tropical cyclone track a natural phenomenon causing excessive torrential rainfall, flash floods, and strong winds H1 H4, Background information The research was conducted in the Lower Sava Valley, a geographic region in eastern Slovenia and part of the Lower Sava statistical region (Figure 3.1), which consists of three major municipalities: Brežice, Krško, Sevnica, and a minor municipality of Kostanjevca na Krki. Since 2015, also municipalities Radeče and Bistrica ob Sotli are part of the Lower Sava statistical region. According to the Statistical Office of the Republic of Slovenia (2014), the Lower Sava statistical region covers an area of 885 km 2, has a population of 70,215 and a population density of 79.3 per km 2. Figure 3.1: The Lower Sava statistical region, Slovenia Source: Statistical Office of the Republic of Slovenia (2014) The region has good logistic infrastructure (railway, highway, airport), with Slovenia s only nuclear power plant (NEK) as one of its main features, along with a thermal wellness centre, 60

83 the Čatež spa resort. The number of enterprises from this region in the respective year was 4,535, and their combined turnover totalled 3,171 million. The geographic region of the Lower Sava Valley (Figure 3.2) lies between the Gorjanci Hills on the south side and the Sava Hills on the north side of the valley. It is riddled with numerous permanent streams as well as intermittent springs and streams. Two major rivers that cross the valley are the Sava and the Krka. The Sava is the longest river in Slovenia with km, flowing through the Lower Sava Valley and entering the Pannonian Basin under the town of Brežice (the International Sava River Basin Commission, 2009). According to the Slovenian Environment Agency (1998), the Sava discharges water from an area of 10,746 km 2. By the size of its basin, the Krka is the largest river that empties into the Sava, taking up 21.4% of the Sava s catchment area. The two rivers join in the municipality of Brežice, near the town of Brežice. Figure 3.2: Geographic region: The Lower Sava Valley, Slovenia Source: Agrež, own research (2015) Between 2010 and 2014, the communities in the Lower Sava Valley experienced four flood events (Figure 3.3) with different severity levels and impacts, owing to different metrological 61

84 and hydrological conditions during the events. There are five flood sources in the valley. In addition to the Sava and the Krka as two major rivers, the streams that carry water from the hills quickly grow into torrents with a threatening power after a few days of continuous rain. The rain itself can cause considerable problems when meteoric water starts to overwhelm the low positioned planes with impermeable soil layers. The communities and the infrastructure located low and near the river can experience groundwater flooding, which usually affects the underground parts of buildings, such as basements, engine rooms, garages, workshops, etc. Figure 3.3: Flood affected communities during the 2012 floods (Dolenjski list, 2012) Source: Agrež, own research (2015) Flood experts believe that the flooding of the Krka in the communities in the municipality of Brežice, which are located within the 8 km area before the confluence with the Sava, is highly dependent on the Sava and its flow rate. High flow rates of the Krka alone represent a threat for the western communities, such as the town of Kostanjevica na Krki in the municipality of Kostanjevica na Krki. However, eastern communities from Cerklje ob Krki to Krška vas may face high water levels but without severe consequences. There are two major reasons behind such hydrological dynamics. Firstly, the town of Kostanjevica na Krki is built on an island which acts as a natural barrier against the Krka flow. At the same time, there is a large 62

85 primeval forest to the northeast of the Kostanjevica island, which acts as a retention area during floods and prevents water from draining out of the area. The second reason lies in the ratio between the flow rates of the Sava and the Krka once they exceed the average rate. The power of the Sava s flow starts to block the Krka s flow, drastically increasing the drainage capacity of the latter. As a result, the Krka floods the communities that are located near its bed and close to the confluence with the Sava H1 H4, Data The data used in the research was obtained from the several different sources. The database administered by the Administration for Civil Protection and Disaster Relief provided the data about 167 entities that faced flood threats during floods and requested an emergency response by dialling the emergency number 112 in the period The data such as municipality, community, address of the entity, and distress type are extracted. Table 3.1: Entity data 1 Municipality Community Entity Altitude Lat. Lon. Brežice Artiče Artiče , , Brežice Boršt Boršt , , Brežice Boršt Boršt10a , , Brežice Boršt Boršt , , Brežice Bračnavas Bračnavas , , Krško Brege Brege , , Krško Brestanica Brestanica , , Krško Brestanica Cestaprvihborcev , , Krško Brestanica Šolskacesta , , Krško Brestanica Šolskacesta , , Krško Brestanica Šolskacesta , , Krško Brestanica Šolskacesta , , Kostanjevicanakrki Koprivnik Koprivnik , , Kostanjevicanakrki Kostanjevicanakrki Kostanjevicanakrki , , Kostanjevicanakrki Kostanjevicanakrki Kostanjevicanakrki , , Kostanjevicanakrki Kostanjevicanakrki Krškacesta , , Kostanjevicanakrki Kostanjevicanakrki Krškacesta , , Kostanjevicanakrki Kostanjevicanakrki Ljubljanskacesta , , Kostanjevicanakrki Kostanjevicanakrki Ljubljanskacesta , , Source: Agrež, own research (2015) Data in the table shows the following attributes for every single entity: altitude, latitude, and longitude (obtained with Google Earth API). In addition, every entity is also ascribed with the 63

86 type of responding teams and the distance of the primary, secondary and tertiary teams from their rescue station to the position of the entity. This data is shown in Table 3.1 (Appendix 1) and Table 3.2 (Appendix 2). I used the same database, administered by the Administration for Civil Protection and Disaster Relief, to extract data about 185 flood distress events during the period of 2010 to Table 3.2: Entity data 2 Distress_type Distance Responder_type Re_distance_1 Re_distance_2 Re_distance_3 Rainfall Volunteers Krka 16 Volunteers Krka 23 Volunteers Krka 50 Volunteers Torrent 25 Volunteers Rainfall Volunteers Torrent 15 Professionals Sava 132 Professionals Torrent 20 Professionals Torrent 73 Volunteers Torrent 65 Professionals Torrent 123 Professionals Rainfall Volunteers Rainfall Volunteers Sava 50 Volunteers Rainfall Volunteers Ground water 10 Volunteers Ground water 640 Volunteers Ground water 336 Volunteers Ground water 32 Volunteers Source: Agrež, own research (2015) The data is shown in Table 3.3 (Appendix 3) with the following attributes: date, entity, activity 1, activity 2, activity 3, activity 4, activity 5 and activity 6, where activities represent measures taken by the responding team that arrived on the site of the emergency. To be able to analyse and compare meteorological and hydrological conditions during the flood events, I extracted the riverflow data (Table 3.4) from the database of the Slovenian Environment Agency, describing river flows during the flood events in the period from 2010 to I compared the river flow measurements from the automatic measuring stations Jesenice na Dolenjskem (Sava River), Hrastnik (Sava River), Veliko Širje (Savinja River) and Podbočje (Krka River). Jesenice na Dolenjskem and Podbočje stations are situated in the 64

87 villages with the same name, which are part of the Lower Sava Valley. Veliko Širje and Hrastnik stations are also situated in places of the same name, but positioned north of the Lower Sava Valley. I included the data of these stations due to the importance of the Sava River levels before entering the Lower Sava Valley and the importance of Savinja River as the Sava influx. Table 3.3: Flood response activities Date Entity KRŠKAVAS Activity 1 water pumping Activity 2 animal rescue Activity 3 collecting animal carcasses Activity 4 people rescue Activity 5 securing the perimeter Activity 6 flood measures BORŠT11 water pumping flood measures other BRAČNAVAS ČATEŽOBSAVI TOPLIŠKACESTA35 water pumping water pumping water pumping flood measures flood measures KRŠKAVAS animal rescue water pumping flood measures GAZICE29 other GAZICE ŠOLSKACESTA KOSTANJEVICANAKRKI ARDROPRIRAKI BREZINA4A Source: Agrež, own research (2015) securing the perimeter removing threats flood measures water pumping water pumping flood measures Similarly, I collected rainfall data from a meteorological database also administrated by the Slovenian Environment Agency which describes rainfall measurements during flood events in the period from 2010 to The data is gathered from the following automatic measurement stations: Bizeljsko, Sromlje, Brege, Smednik, Kostanjevica na Krki, Planina and Cerklje ob Krki and presented in Table 3.5. All mentioned stations are situated in the Lower Sava Valley, except for Smednik, which is situated west of the valley, yet rainfall in that area contributes to the Krka River water level. 65

88 Table 3.4: Hydrological data River flow date Jesenice na dol. Podobočje Veliko Širje Hrastnik Source: Agrež, own research (2015) Table 3.5: Meteorological data Rainfall date Bizeljsko Sromlje Brege Smednik Kostanjevica Planina Cerklje

89 Source: Agrež, own research (2015) 3.2 Testing Hypotesis 1 In the first hypotesis, I tried to determine whether loosely coupled systems are built with a process architecture that can be defined and assessed. I designed a virtual model of the flood response system as an example of a loosely coupled system from the early stage when its widgets are active only as individual organizations with no active role in the system, until the state when the system is fully operative and all its widgets assume an specially delegated role Methodology To be able to determine whether the definition and assessment of such a system is possible, I used the methodological approach that is usually used when assessing conventional business systems. First, I gathered the data that describe the processes and entities involved in the system, and then designed process pattern models, using an adjusted table presented in Table 3.6 (Appendix 4) and Table 3.7 (Appendix 5). The adjusted TAD activity table integrates part of the activity table and part of the table with characteristics of the Tabular Application Development (TAD) methodology (Damij, 2000). 67

90 Table 3.6: Adjusted TAD activity table 1 Activity Duration Entity Pattern Activity no. (min) ReCO 1 1 information monitoring --- ReCO 1 2 information analysis 5 ReCO 1 3 standby --- ReCO 2 4 receive call 3 czcommand 1 33 standby --- czcommand 1 34 regular office work --- czcommand 2 35 activation --- czcommand 2 36 command centre establishment 60 czrespond 1 78 receive deployment task 3 czrespond 1 79 move to the distress entity --- czrespond 2 80 conduct a field survey 10 czrespond 2 81 report to the CZ command 3 firecommand 1 82 not active --- firecommand 1 83 standby --- firecommand 2 84 activation 10 firerespond 1 99 standby --- firerespond receive reco activation 3 firerespond gathering 10 roadcompany standby 1 roadcompany regular standby work --- roadcompany receive road block information 1 geotec standby 1 geotec regular standby work --- geotec receive landslide information 1 Source: Agrež, own research (2015) Further, I designed a simulation based mostly on the script coded in the R programming environment supported by the Weka and Rapid miner datamining tools. The simulation (Figure 3.4) is composed of four inputs: river flow data, rainfall data, profile classes that determine endangerment, and entities taken from the entity table. It integrates the C4.5 data 68

91 mining algorithm (Quinlan, 1993) and the Slovenian standard operating procedures for disaster response, designed as a decision tree. Table 3.7: Adjusted TAD activity table 2 No. Activity U P h m U P h m U P h m U P h m 1 Wake up Morning activities Daytime activities / / / / / 3 / / / / / / 5 / Evening activities Get ready for sleep Go to sleep / / / / 31 5 / / / 10 / / / / 5 0 / / / 5 / / / / Source: Agrež, own research (2015) The first input consist of merged meteorological and hydrological data, used as a trigger for establishing command authorities of the local association of voluntary fire brigades and the local Civil Protection Service. The second input consists of classification rules based on which profile classes for the included entities (households from the flood-endangered areas) are formed. The third input includes the entities selected to participate in the simulation run. Entities data is gathered in the TAD entity table. The final input consists of adjusted TAD activity table with activity flow notation of entities daily activities. The simulation run is executed in three stages. In the first stage, entities are allocated into classes specifying during which meteorological and hydrological conditions respective entities will become flood-endangered. Further, they conduct their daily activities according to the TAD activity table until the rainfall and river status trigger an alert. The second stage begins with the entities decision to call an emergency centre or the Civil Protection Service and ask for a response, depending on the flood threat they are exposed to. At the same time, command authorities of local disaster response services are being established with a potentially critical rise of the river flow rates. The operator at the distress call centre or the civil protection officer who has received the call decides on the most adequate responding authority, depending on the situation, and forwards the information to the competent response service. As part of the third stage of response services, either a local Civil Protection Service or one of local fire brigades respond to the distress call in line with their standard operating 69

92 procedures. In the research, the term standard operating procedures refers to official documents prepared by the local government and response services which must be consistent with the national standard operating procedures. Figure 3.4: Structure of the simulation Source: Agrež, own research (2015) 70

93 3.2.2 Results The primary logic which runs the first and the second stages of the simulation is based on the decision tree classification with precision of 94.21% (Table 3.8 and Table 3.9). Table 3.8: Classification confusion matrix a b c d e f a = groundwater b = Krka c = rainfall d = Sava e = torrent f = no Source: Agrež, own research (2015) The simulation was executed using the original data gathered from the four flood events as the input, subjected to decision rules. The red outline in Figure 3.3 represents the area (longitude and latitude) in which the respective entities are located. Within the simulation run, all entities were exposed to meteorological and hydrological conditions as recorded during the actual flood events. Such simulation scenario enabled a comparison of data, representing the overall flood response process and the classification-generated simulation output. The simulation output of the first and the second stages included entities allocated into several classes according to the expected flood threat. The entities were merged into communities based on their geolocation and the geographical borders of local communities. Table 3.9: Detailed accuracy by class TPRate FPRate Precision Recall F-Measure ROCArea Class Groundwater Krka Rainfall Sava Torrent No Source: Agrež, own research (2015) A comparison of Pareto charts with the original data and simulation output (Figure 3.5) reveals a satisfactory matching of simulation results with the original data. 71

94 Figure 3.5: Pareto chart comparison of the original data and simulation output Source: Agrež, own research (2015) The nearest matching was identified between real data and simulation output of the meteorological and hydrological conditions that represented no flood threat to the communities. A comparison regarding ground water, the Sava, and rainfall as potential flood 72

95 threats indicated that there was a minor derogation between both data types, with the simulation run occasionally classifying more entities as endangered compared to the classification based on real data. A comparison regarding the communities endangered by the Krka revealed a similar derogation, although in the opposite way. According to the simulation, several entities were classified as less endangered compared to real data classification. The simulation output of the third stage was based on previously allocated entities. It comprised the information such as which responding organization engaged in the emergency response, how many times, and in which community. The output chart shows the response process workload of the emergency responders. In the bubble chart, the X-axis displays communities and the Y-axis displays responding units. Bubble sizes and bubble colour indicate responding frequency and the entity in distress, respectively. The criterion in selecting the units of the primary, secondary, and tertiary response forces was the distance between the entity in distress and a particular unit. The primary response force during flood events is shown in Figure 3.6. Figure 3.6: Response process workload of the primary response force Source: Agrež, own research (2015) The responding units can clearly be related to the community. The closest responders reacted to a distress within the community. The most notable derogation was identified with the Professional Fire Brigade (PGE), which responded also out of the community where it was geographically located. 73

96 In the case of the secondary response force (Figure 3.7), the response process workload is geographically more dispersed. Figure 3.7: Response process workload of the secondary response force Source: Agrež, own research (2015) A simulation revealed that the workload of the units engaged in the response process went beyond the associated communities and their primary response scope. An even more dispersed response would be present in the case of tertiary response force engagement (Figure 8). Compared to the primary response force, the tertiary force lacks strong community-based relations, meaning that the responding units would service entities in distress on a much broader scale than their normal scope. Figure 3.8: Response process workload of the tertiary response force Source: Agrež, own research (2015) 74

97 With the obtained results, H1 could be confirmed. Not only was I able to model process architectures of the chosen loosely coupled system, I was also capable of transforming the system into a dynamic process simulation which provided a satisfactory virtual replica of the real world system. Based on the simulation, I was able to get the insight into how different scenarios would generally affect outputs of different process architectures. 3.3 Testing Hypothesis 2 The second hypothesis stated that it is possible to map and assess community knowledge. In order to verify whether such an analytical approach is possible, I designed a learning evaluation environment which provided an insight about community learning success rates and also created the basis for further knowledge mapping Methodology At this stage, I began with the semi-structured interviews. I interviewed people included in the households which are represented as entities in the database. A total of 56 respondents were selected from flood-endangered entities. Participant recruitment criteria were: active involvement in at least one flood event, residing in one of the households affected by the detected distress sources, and available open access to all the information about the ongoing situation, along with their family members and other members of the household. Table 3.10: Semi-structured interview framework Before the first event During the first event After the first event Before the following event During the following event After the following event Risk awareness Disaster risk information Household response Knowledge source x x x x x x x x x x x x x x x x x x x x x x x x Source: Agrež, own research (2015) 75

98 The respondents included young adults and other adults (34 male, 22 female) with the mean age of and standard deviation of The aim of the interviews was to detect how the entities perceived their own learning about the floods, the circumstances, and response actions of responsible services. I used an interview framework in a matrix form shown in Table 3.10 and analysed the results with a paired t-test. In the third and final step, I first compared rainfall and river flow attributes recorded in the four flood events with the multivariate analysis of variance, to be able to find a similarity of the flood events through flow rate and rainfall measurements, using Wilk s lambda as a probability distribution. I compared the data by year of occurrence, trying to discover which attribute (flow rate in the first case and rainfall quantity in the second) influenced the difference most significantly. Based on the events similarity and learning the rules from the third step, I further designed a fuzzy system (Hornik and Meyer, 2009) presented in Figure 3.9, which served as a tool for measuring the learning performance of flood-endangered communities. Figure 3.9: Designed learning evaluation fuzzy system Source: Agrež, own research (2015) Finally, I selected entities from the communities in which learning performance was measured on the scale from average to excellent, and calculated the correlations among situational and 76

99 response attributes to be able to determine which general learning rules applied to learning communities Results By conducting semi-structured interviews, I was able to design the interview data nominal scale shown in Table Table 3.11: Interview data nominal scale Detected variants Field of inquiry Disaster risk No Present Moderate Influential High awareness Disaster risk information source Media Observation 77 Government agencies Public services Household response Observation Preparedness Response Recovery Past Knowledge source experiences Source: Agrež, own research (2015) Ongoing experience / Risk reduction / / / In the above table, values from 1 to 5 are assigned to the four categories which I was investigating during the interviews. Every value stands for the detected variant identified among all the answers. I used the scale to convert the gathered answers into nominal data presented in Table 3.12 (Appendix 6), Table 3.13 (Appendix 7) and Table 3.14 (Appendix 8). Table 3.12: Quantitative results of the semi-structured interview 1 # Disaster risk awareness Before the first flood Disaster risk information (type and source) Household response Emerged knowledge (type and source) Disaster risk awareness During the first flood Disaster risk information (type and source) Household response Emerged knowledge (type and source Source: Agrež, own research (2015)

100 Table 3.13: Quantitative results of the semi-structured interview 2 After the first flood Before the second flood # Disaster risk awareness Disaster risk information (type and source) Household response Emerged knowledge (type and source Disaster risk awareness Disaster risk information (type and source) Household response Emerged knowledge (type and source Source: Agrež, own research (2015) Table 3.14: Quantitative results of the semi-structured interview 3 During the second flood After the second flood # Disaster risk awareness Disaster risk information (type and source) Household response Emerged knowledge (type and source) Disaster risk awareness Disaster risk information (type and source) Household response Emerged knowledge (type and source) Source: Agrež, own research (2015) I continued with the paired t-test (Table 3.15) to discover possible similarities between the fields of inquiry before, during, and after the first and second flood events. The results of the test revealed that the only significant difference discovered was the disaster risk awareness before the first and the second flood events. Other fields of inquiry, including knowledge source, were not significantly different. An analysis of the interview data, including the t-test, 78

101 revealed that during and after the floods, households in the flood-affected communities used mostly real-time experiential learning, based on the ongoing situation. Before the floods, their knowledge source was mostly past experiential learning. Table 3.15: Mean, standard deviation values and paired t-test Paired t-test Field of inquiry First flood event Second flood event t p Interpretation Before During After Disaster risk awareness Disaster risk information Household response Knowledge source Disaster risk awareness Disaster risk information Household response Knowledge source Disaster risk awareness Disaster risk information Household response Knowledge source Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Mean SD Source: Agrež, own research (2015) Significant Not significant Not significant Not significant Not significant Not significant 0 1 Not significant Not significant 0 1 Not significant Not significant Not significant Not significant The interviews also provided information on the similarity between the respective flood events in terms of their implications for the communities. Information similarity regarding implications was based on the distress source affecting the community. To get an objective comparison, I conducted a multivariate analysis of variance using flow rate data and rainfall data. Table 3.16 compares Wilk s Lambda values from flow rate multivariate analysis of variance testing. The results of the comparison of flood events for the years 2012 vs and 2013 vs show a significant difference only for the Podbočje station flow rate. Furthermore, a nearly significant difference was identified with regard to the Hrastnik station 79

102 flow rates for the years 2010 vs and 2012 vs There was no significant difference in the comparisons between 2010 vs and 2010 vs Figure 3.10 provides an overview of flow rate measurements during all four flood events, with recorded differences in the river flow rates. Table 3.16: River flow data, multivariate analysis of variance Jesenice station, flow rate Podbočje station, flow rate Širje station, flow rate Hrastnik station, flow rate Year p Wilk`s λ F-value Wilk`s λ F-value Wilk`s λ F- value Wilk`s λ F-value e Source: Agrež, own research (2015) Figure 3.10: Flow rate measurements during flood events Source: Agrež, own research (2015) 80

103 Table 3.17: Rainfall data, multivariate analysis of variance Bizeljsko Sromlje Brege Smednik Year p Wilk`s Wilk`s Wilk`s F-value F-value λ λ λ F-value Wilk`s λ F-value Kostanjevica Planina Cerklje year p Wilk`s λ F-value Wilk`s λ F-value Wilk`s λ F-value Source: Agrež, own research (2015) Based on the distress types affecting the flood-endangered communities, I also included rainfall measurements in the research. Rain is, as opposed to river flow rate, the main source of distress situations caused by torrents and meteoric water. I conducted a multivariate analysis of variance using rainfall measurements from the meteorological stations close to the 81

104 flood-affected communities. As can be seen from Table 3.17, no rainfall measurement was identified which would make a clear distinction among the respective flood events. Therefore, I was able to determine, relying also on the measurement plot in Figure 3.11, that the most notable similarities were observed between the events in 2010 vs and 2013 vs Figure 3.11: Rainfall measurements during flood events Source: Agrež, own research (2015) I developed the criteria for fuzzy evaluation of community learning based on the insights into experiential learning and risk awareness from the t-test and event similarities from both multivariate analyses of variance. At the first stage of the evaluation, I divided the complete set into five subsets, based on the distress type. The structure of every subset was A = {B, C, D, E}, where A represents the distress type subset, B the subset of detected distress during the first flood event, C the subset of detected distress during the second flood event, D the subset of primary activities during flood events, and E the subset of secondary activities during flood events. The following stage of evaluation included the categorical evaluation in which learning performance of the communities was measured using the condition of the same distress source during both flood events, and following five criteria: x1 B x2 C L1 = {1}, distress is detected during both flood events; x1 B L2 = {1}, distress is detected only during the first flood event; x2 C L3 = {1}, distress is detected only during the second flood event; 82

105 x3 D` x3 D`` L4 = {1}, primary response activities are the same during both flood events; x4 E` x4 E`` L5 = {1}, secondary response activities are the same during both flood events. The subset X = {x1, x2, x3, x4} represents data typical for a single entity (communities are composed of entities, which are in reality flood endangered households), while the subset L = {L1, L2, L3, L4, L5} represents its learning grades. To be able to place the grades of the entities into the community perspective, I summed them into community subsets CS n = n { L1n L1, n L2n L2, n L3n L3, n L4n L4, n L5} CS n ={CL1n, CL2n, CL3n, CL4n, CL5n} and further weighted them with the weighting rules presented in Table Table 3.18: Weighting criteria L5n CL1n CL2n CL3n CL4n CL5n Value ranks Value Weight Value Weight Value Weight Value Weight Value Weight High rank 3 8 <=16 8 <=16 8 >10 15 >1 15 Middle rank 2 4 < 6 4 < Low rank < Source: Agrež, own research (2015) I applied the weighted subsets to the previously designed fuzzy system and obtained the results which revealed the most successful learning communities (Figure 3.12). Out of the total 59 communities, only 10 demonstrated a learning success that was above poor. Figure 3.12: Fuzzy system plot of the most successful learning communities Source: Agrež, own research (2015) 83

106 The results of the second part of the research confirmed that community knowledge can be assessed, even though it depends on the learning process which occurs mostly within the emergency-based loosely coupled system. At the same time, I was able to determine that community learning which occurred when the system was active and in operating mode was experiential learning and that the acquired knowledge was of tacit nature. No systematic learning measures were detected that would provide local communities with the necessary flood protection knowledge and consequently make the communities more self-reliant and flood resilient. Therefore, the knowledge map which could provide an insight into arranged, systematic and explicit knowledge cannot be built on the basis of detected community learning within the observed loosely coupled system, even though I designed the knowledge correlation framework which showed how different knowledge sources correlate among each other, observing them from the perspective of all communities included into the research. Figure 3.13: Correlations of attribute values, typical for entities from successful learning communities Source: Agrež, own research (2015) 84

107 The correlation analysis of selected attributes describing the most successful learning communities, presented in Figure 3.13, revealed the highest correlation between altitude and community, distance from distress source and entity address, distress type and community, and distress type and distance from distress source. In addition, logical correlations exist between community and entity address as well as entity address and its altitude, which I neglected due to irrelevance. The correlations revealed a learning frame within which experiential learning of the communities took place. Situational knowledge sources, connecting distress type, distance from the distress source, and altitude of the households with the community were typically more exposed than the connections among response units, response activities, and communities. In the scope of the assessed loosely coupled system and experiential learning employed by communities during flood events, I can undoubtedly confirm H2 to be true. It is possible to assess and map community knowledge in the loosely coupled system Interviews with civil protection representatives An interview with Mr. Roman Zakšek, the representative (Mayor s counsellor and deputy commander) of the local civil protection services, revealed a similar assessment of the flood endangered local communities when seen as learners, knowledge hubs or even knowledge sharing initiatives. He finds that, in general, local communities are not aware of natural disaster threat possibilities. Residents almost never respond when invited to local presentations of natural disaster standard operating procedures specifically designed for the local area. Among natural disasters that could affect the area, flood events influence only those villages that are directly affected by the water threat. Villages situated close to the affected ones, but not directly endangered by the floods, mostly feel no direct influence and consequently those households and villagers feel no need to respond or take action. They do not learn about threatening situations from those who are directly affected. An interesting trend is also the fact that households that call the local civil protection services for support during flood events, are the households included in the local communities that face a direct flood threat by regular, annual high waters. Households from the communities which got affected only by the most fierce flood event in the assessed period did not contact their local civil protection unit when faced with additional flood threats. Situational reactions, based on the current experience with the 85

108 ongoing factors, are more common than prevention-oriented thinking and forward-oriented decision making. Local civil protection services try to inform and educate the residents of flood endangered local communities so that they would be better prepared for future flood events, which seem to be practically inevitable. The response to such learning and information sharing initiatives is negligible. Flyers distributed among households who are the first to face a flood threat encourage no extra residents to contact the local civil protection service to share information about the incoming flood threat or to ask for support or instructions on how to respond. At the same time, only the responding organization in the system which emerges during flood events attended a locally organized seminar dedicated to flood prevention and protection. Residents or formal representatives of flood endangered local communities did not attend the seminar, even though they were invited. The local civil protection service desires to educate residents and local communities to be able to take over response initiatives, and to increase their self-reliance and flood resilience. To this end, Mayor s advisor for the field of civil protection designed a brief, interesting and motivational course for emergency first aid CPR procedures. During these courses, the civil protection members meet the local community members, share some basic knowledge, and try to motivate the residents to join the civil protection teams in order to receive additional knowledge, and, most of all, awareness about locally present natural disaster threats. Going into the community to talk to the people is of crucial importance for the villagers and other residents in order to rebut the thinking where their own protection and safety is entrusted to responding structures, which are often perceived by the general public within the communities as something distant, closed and hardly accessible. Another interview was conducted with the Head of the regional office of the Administration for Civil Protection and Disaster Relief, Ms. Zdenka Močnik. Her role is managing the regional office in the fields of monitoring and study of natural hazards and natural disasters, planning of preventive activities, providing the necessary information, issuing warnings and alarms of direct threats, providing instructions for protection, rescue and aid. The regional office also organizes the regional civil protection and establishes and maintains other forms of standby protection, rescue and aid, mobilization and activation of regional and national forces and means of protection, rescue and aid, ordering and 86

109 implementation of protective measures, rescue and aid assessment for damage caused by a natural disaster. She emphasized that continuous threat awareness exists within the communities if flood events occur constantly, on an annual basis. These communities face flood threats even when almost everyone else is still safe. Conversely, communities that faced a direct threat only once and then remained unharmed, slowly neglect their concerns about efficient flood preparedness. Such communities learn and react according to the ongoing situation. Their situational awareness and situational learning is strong, but after the experiences slowly fade out, the residents get back to their daily routines with no special thought on possible further flood threats. In some specific situations, even households constantly facing flood threats find no real option of establishing efficient flood protection due to the national regulations that legally prohibit interventions into riverbeds and swamps that used to represent partial solutions to flooding issues. However, inhabitants follow and consider meteorological and hydrological warnings with much more concern than in the past. Such warnings and forecasts have become accurate and widely accessible to the local communities in general. Explicit knowledge hubs in the local communities are fire brigades and firefighters, volunteers. They are constantly attending educational courses and trainings organized by the Training Centre for Civil Protection and Disaster Relief of the Republic of Slovenia. The Centre provides basic educational programs as well as advanced courses, based on the state of the art and relevant case studies. It must be remembered, however, that the educational approach is focused on responding structures and consequently does not include community members, villagers, and residents who are not part of responding organizations. Therefore, people rely on the fire brigade and the civil protection to serve them when needed, including a response during flood events. Motivation of the general public to learn about floods and other natural disasters and to develop knowledge-based community flood protection is, in general, not at the level where concrete measures could be taken. Even urgent topics, such as nuclear safety in the region, do not motivate the local communities to attend presentations where they could at least ask provocative questions. Professional public, in this case the civil protection members and fire fighters, are attending such presentations, which also attests to the tendency among the communities to shift the disaster response burden mostly on responding organizations. 87

110 However, some bright examples in the form of individual households do exist. There are households facing constant flooding who learn through the response process and could easily use the same equipment as the fire brigade. By providing water pumps and learning how to use them, these few households raised their flood resilience without creating an extra workload for the responding organizations. In order to encourage a wider rise in flood resilience and self-reliance awareness, I would need to address the as-is state systematically and professionally, using the state of the art knowledge and examples from other countries already proven to work under similar conditions. Interview with Mr. Branko Dervodel, Deputy Director General of the Administration for Civil Protection and Disaster Relief, provided a general overview of the assessed issue. Mr. Dervodel sees Slovenia as very exposed to natural disasters, therefore some communities are regularly facing threatening situations, among which floods are the most frequent. Such individuals and communities have a tendency to take care of their own flood protection. Alternatively, there are individuals and communities that look at disaster events from a more opportunistic angle. After having been faced with a natural disaster event and not exposed to a series of similar events through the years, their memories slowly fade and they neglect the measures aimed at improving their disaster resilience. There is also a distinct difference between rural and urban communities. The so-called rural natives, who had lived in the same place for a generation, know very well what kind of threats to expect and how to protect themselves. They built a high resilience and are self-reliant, while urban communities mostly helplessly observe the progress of the emergency and demand help from responding organizations. Administration for Civil Protection and Disaster Relief treats all regions and residents equally. It is no longer the case that flood events would come as a surprise due to high-quality meteorological services and a well spread alarming and informing system. Those individuals and communities who take disaster warnings seriously also have a tendency to ensure sufficient material and technical means for their own protection. Households prepare to withstand flood events together with fire brigades. There are also communities which had a much higher awareness about natural disasters and flood events in the past, compared to the state recorded for this past decade. In trying to identify the interest of communities for gaining new knowledge about natural disaster and flood protection, I detected different perspectives. Big promotional and 88

111 information sharing events, such as the Rescue and Protection Days, organized in frequent intervals in larger Slovenian cities, are well visited, as are creative contests for children. Moreover, educational topics covered in schools are well received. The question arises of whether visitors and participants learn about how well equipped, staffed and trained the response system is, and whether they use the knowledge and information gained to raise the awareness about the importance of self-reliance during flood events and other natural disasters. A perceptible difference exists between municipalities where preparedness varies due to their financial capabilities, knowledge and influences of key decision makers. Sometimes they neglect the fact that, in the first moments of distress, a person is usually left to their own knowledge and skills. Officials try to encourage learning by addressing the population trough leaflets, social networks, mobile applications and by organizing promotional and educational events. In general, the investigated communities are resilient to natural disasters. It is hard to tell whether this is due to frequent experiences with floods or other events, or due to information and knowledge sharing. There are some best practice examples. Ljubljana municipality is one of them, as it takes good care of its disaster protection mechanism. There is also the Kamnik municipality, which obtained an ISDR (International Strategy for Disaster Reduction) certificate, proving its excellence in natural disaster preparedness and reduction. Knowledge exchange at the operative levels is mostly present in firefighters who are willing to share new and specific knowledge obtained from specific interventions with other units. Municipalities must plan and consequently foresee their knowledge needs, based on detected exposure to natural disaster thrests. The role of the Administration for Civil Protection and Disaster Relief is to support municipalities when needed and to ensure that their preparedness plans and standard operating procedures are in line with national plans and procedures. Disaster response in local communities is based on the local firefighting organizations. When they can efficiently response independently, they will take over the whole workload and request additional support only when needed. Our final interview reveals the view of the Higher Counsellor in the Department for Protection, Rescue and Civil Defence of the Ljubljana municipality, Mr. Julij Jeraj. In his work to date, he identified a clear distinction between families who had lived in a local community for generations and newcomers. Families that passed their homes from generation to generation are used to living with natural disasters, and can be usually found in rural areas. 89

112 They will be able to take care of themselves because they have developed flood protection measures, using both convenient and commercial solutions. Newcomers are those who had moved from an urban community over the past years. Their lifestyle directs them to meet their needs by looking for services and product providers in those segments in which their need appears. It is the same during flood emergencies: they demand the help of responding organizations, even though they could at least try to establish a self-protecting environment. It is an interesting phenomenon that with time, the response workload of responsible organizations combined with a generally unchanged frequency of natural disasters, visibly increased the number of support requests. The interest of the local community members to gain new knowledge on natural disasters and flood protection is, in general, at an unsatisfactory level. The municipality prepared leaflets, organized events, and set up a web page, but only a handful of residents showed any interest. What is more, the residents of flood-endangered communities claimed that the civil protection did not provide them with waterproof boots, revealing poor awareness about the importance of self-reliance and flood resilience. It also happened that a group of people used high water for wakeboarding. They caused waves, raising the critical level so much that it crossed doorsteps and entered the houses with incoming waves. The municipality itself organized training for firefighters so that they would be able to respond to flood events. It was the municipality s initiative even though not all firefighting brigades were interested in obtaining this knowledge. In contrast, the humanitarian organization Slovene Philanthropy from Ljubljana offered the help of their volunteers to support the response during flood events. Their help was turned down by the municipality, due to doubts over how these volunteers could enhance the established response system, over the level of knowledge, their insurance coverage, their extent of self-reliance, and due to calls for avoiding unnecessary confusion when masses of people try to help with no knowledge, being ineffective. It would be possible to include groups of volunteers into the response system, but only following predefined terms and an agreement of how their support should be implemented. There are individuals who want to increase their awareness of natural disasters and flood protection. Sooner or later, they will ask for additional sources of knowledge or materials on how to improve their safety and resilience. On the other hand, there is the poorly developed systemic knowledge management with a lack of methodology on how to select key 90

113 individuals in municipalities or local communities to manage disaster protection. There is no career development, no proper guidance, consequently making it hard for the disaster response system to be efficient as a whole. 3.4 Testing Hypothesis 3 The third hypothesis stated that community knowledge can influence processes in loosely coupled systems. In order to verify whether the H3 is true of false, I had to design a way to influence processes within the operating loosely coupled system, using additional knowledge not yet introduced to the system, and entities that form it. After the learning-based influence, I compared the loosely coupled system process output with the original outputs to check whether there were any notable differences Methodology I designed an optimization algorithm (Figure 3.14) that simulates additional systemic community learning in addition to experiential learning. I collected the basis for optimization rules from the Resolution of National Security Strategy of the Republic of Slovenia (National Assembly of the Republic of Slovenia, 2010) and the Resolution of National Program of Protection against Natural and other Disasters for the period (National Assembly of the Republic of Slovenia, 2009). The optimization algorithm foresees several educational prevention measures in order to increase resilience and self-reliance of flood-endangered communities: - Improvement of general flood preparedness; - Systematic modernization of training programs; - New educational programs for pre-school and school children; - Encouragement of measures that raise prevention, protection and security awareness. I applied the simulation of educational prevention measures on the flood event data, including the following process dimensions: communication time, travel distance, number of process architectures, number of process patterns, number of activities, number of entities in distress, total number of executed standard operating procedures, and number of different standard operating procedures during one event. I considered the severity level of each assessed flood event. According to recorded occurred damage, scope of the flooding, and response force 91

114 activity workload, I determined the event horizon when an individual household becomes incapable of providing self-reliant flood protection and must request for an emergency response. Figure 3.14: Process optimization algorithm design Source: Agrež, own research (2015) Results In the final stage, I simulated learning and prevention optimization of the flood-endangered communities. I considered a comprehensive set of predefined optimization measures to be able to predict self-reliant community flood protection. Table 3.19 shows the significance of differences between the as-is response process state and to-be response process state. Paired t- tests of both states process output data revealed significant differences for every category, except respondents travel distance. To be able to interpret the results, assessed process data categories must be addressed. Process optimization reduces: communication time to 66.67%, responders travel distance to 43.28%, number of process architectures to 61.6%, number of process patterns to 68.33%, number of activities to 66.13%, number of entities in distress to 60.73%, the total number of executed standard operating procedures to 55.81%, and the number of different standard operating procedures to 63.74%. 92

115 Table 3.19: Mean values, standard deviation values and paired t-test of AS-IS and TO-BE process states Paired t-test Field of inquiry AS-IS process state TO-BE process state t p Interpretation Mean Communication time SD Mean Respondents` travel extent SD Number of process Mean architectures SD Mean Number of process patterns SD Mean Number of activities SD Mean Number of entities in distress SD Total number of executed Mean standard operating procedures SD Number of different standard Mean operating procedures SD Source: Agrež, own research (2015) Significant Not significant Significant Significant Significant Significant Significant Significant Even though the t-test showed the difference between as-is and to-be states of responders travel extent as not being significant, the optimization algorithm reduced the extent of travel by 56.72%, which is the highest optimization rate within the simulation. The insignificance cannot be attributed to the travel extend reduction, but to the comparison of the data distribution in the as-is and to-be states of the process output. This phenomenon can be attributed on the one hand to the fact that the optimization algorithm reduced the travel extent mainly due to fewer communities which requested emergency assistance in the to-be state. On the other hand, the significant differences arise mainly from the process changes in the communities which were not excluded from the to-be state, while their data distribution significantly changed. The optimization algorithm affected every included process dimension. Figure 3.15 represents dependencies between all pairs of process dimensions that were shown to be significantly different when in as-is or to-be state. The colour blue indicates the as-is state, and the colour red indicates the to-be state. The plot clearly reveals similar dependencies between dimensions in both states, while dispersion clearly reduces in the to-be state, revealing 93

116 successful process optimization. When compared to other dimensions, similar dependencies between the following dimensions also become evident: number of activities and entities in distress; number of process architectures and process patterns; total number of executed standard operating procedures and number of different standard operating procedures. Detected similarities could represent further optimization possibilities through the yet unclear interdimensional connections. Figure 3.15: Scatter matrix of AS-IS and TO-BE process states data Source: Agrež, own research (2015) Based on the obtained results, H3 can clearly be confirmed. Results evidently revealed that integration of new knowledge into the assessed loosely coupled system importantly influences the operation of the system itself, which becomes more efficient when using originally threatened entities and an auxiliary work source that relieves the primary response force and creates a higher capacity of response force activity focus. 94

117 3.5 Testing Hypothesis 4 The fourth hypothesis stated that community learning in the loosely coupled system is a mutual process. For testing the hypothesis, business process management methods were used to model the loosely coupled system and to further simulate it in order to conduct a statistical analysis of the simulation results Methodology Here, I first designed a process model with the TAD activity table and upgraded it with the knowledge module. The model reveals several process scenarios which differ according to activity flow and the knowledge exchange among flood endangered entities and organizations responsible for the emergency response. In the second step, I translated the process model into the dynamic simulation, using the igrafx simulation tool. I executed a simulation of 33 process scenarios. Further, I extracted from the simulation results the overall number of transactions (transitions between activities), the number of positive decisions and the number of negative decisions of influential decision makers. Moreover, I counted the number of knowledge links (executed transactions which contributed to the emergence of new knowledge), directed from the flood endangered entity towards responding entities (red links) and in the opposite direction (blue links). Using descriptive statistics, I presented the basic statistical insights into the obtained data. Finally, I conducted several different analyses of previously conducted transaction counts data. A principal component analysis revealed the most influential of the observed variables. Multiple correspondence analysis revealed three major groups of process scenarios. Using principal component analysis and hierarchical clustering based on principal component analysis, I obtained similar results indicating three major types of process scenarios. To be able to get an insight into which of the included variables influenced knowledge spread, the chi-squared test was performed. I compared the transaction count of every entity with the number of corresponding knowledge links within the process scenarios. 95

118 3.5.2 Results The flood event response process model shown in Figure 3.16 is composed of 23 activities and 8 entities. The model includes all key decision points which influence the course of activity flow and knowledge exchange. With conversion into a discrete simulation and execution of 33 predefined scenarios, I was able to collect activity transaction data linked with not only the key entities, but also their decision making. Figure 3.16: Flood event response process model Source: Agrež, own research (2015) 96

119 Scenario Community member transactions Community member positive decisions Community member negative decisions Distress call centre transactions Distress call centre positive decisions Distress call centre negative decisions CP command transactions CP command positive decisions CP command negative decisions CP response transactions Fire brigade command transactions Fire brigade command positive decisions Fire brigade command negative decisions Fire brigade transactions Road company transactions Geo-technician transactions Table 3.20: Transaction count Transaction count Source: Agrež, own research (2015) 97

120 The process simulation yielded results in the form of transactions. I extracted the transaction count for each scenario and gathered them in a joint database, shown in Table In order to produce a general assessment of transaction dynamics, I calculated the means and standard deviations of the transaction count, where I included the key entities of the flood event response process and their decision making activities (Table 3.21). Table 3.21: Transaction count, mean values and standard deviation Transactions Mean SD Community member transactions Community member positive decisions Community member negative decisions Distress call centre transactions Distress call centre positive decisions Distress call centre negative decisions CP command transactions CP command positive decisions CP command negative decisions CP response transactions Fire brigade command transactions Fire brigade command positive decisions Fire brigade command negative decisions Fire brigade transactions Road company transactions Geo-technician transactions Number of red knowledge links Number of blue knowledge links Number of knowledge links Source: Agrež, own research (2015) Figure 3 shows a comparison between proportions of mean values and standard deviation values. The comparison provides an interesting insight of how the entities and their decisions differ throughout the 33 simulated scenarios. Both Table 3.21 and Figure 3.17 state the highest average number of community member transactions with a small standard deviation. The closest ratio is found only with the smaller transaction count average value, and higher standard deviation value, by the community member positive decision transaction count, 98

121 followed by the community member negative decision transaction count. A similar pattern follows all key entities with their decision making transactions. Figure 3.17: Comparison between proportions of mean values and standard deviation values geotechnician transactions road company trasactions fire brigade transactions fire brigade command negative decisions fire brigade command positive decisions fire brigade command transactions CP response transactions CP command negative decisions CP command positive decisions CP command transactions distress call center negative decisions distress call center positive decisions distress call center transactions community member negative decisions community member positive decisions community member transactions 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% mean SD Source: Agrež, own research (2015) In addition to the transaction count focused on entities and their decision making, I also conducted transactions which directly indicate the knowledge exchange between two groups of entities: costumers (flood endangered entities) and providers (flood responding organizations), as shown in Figure Red knowledge links are named transactions and indicate a knowledge exchange in the direction from costumers to providers. Blue knowledge links are transactions which indicate a knowledge exchange in the direction from providers to customers. Within the set of 33 simulated scenarios, I was able to detect a minor group of 6 scenarios, where knowledge exchange runs only in the direction provider to customer. The rest 27 scenarios demonstrate knowledge exchange in both directions. A graphical representation also indicates several sets of scenarios with similar characteristics, referring to knowledge exchange. 99

122 Figure 3.18: Comparison of knowledge exchange transactions number of red knowledge links number ok blue knowledge links Source: Agrež, own research (2015) Multiple correspondence analysis was conducted to get an insight into the possible subgroups within the set of 33 process scenarios. The multiple correspondence analysis (Figure 3.19) revealed three major groups of process scenarios. Figure 3.19: Multiple correspondence analysis plot Source: Agrež, own research (2015) 100

123 Even though the multiple correspondence analysis revealed subdivisions of the dataset, I was unable to understand which attributes contributed most to detected subgroups. Therefore, the principal component analysis was used. Principal components analysis plot (Figure 3.20) shows that among all the principal components, the first three cover a major proportion of the variance. Figure 3.20: Principal component analysis plot Source: Agrež, own research (2015) Table 3.22 shows the standard deviation and proportion of variance of the detected principal components, and reveals that Principal Components 1, 2 and 3 represent 44.97%, 20.43% and 19.02% of the variance, respectively. All other components follow with 5% and less. 101

124 Table 3.22: Importance of components PRINCIPAL COMPONENTS Standard deviation Proportion of variance Cumulative proportion Source: Agrež, own research (2015) Principal component analysis revealed the most influential attributes considering the process scenarios. Table 3.23 shows these attributes placed within the principal component frames. Table 3.23: Attributes placed in the framework of principal components Transactions PC1 PC2 PC3 PC4 community.member.transactions community.member.positive.decisions community.member.negative.decisions distress.call.center.transactions distress.call.center.positive.decisions distress.call.center.negative.decisions CZ.command.positive.decisions CZ.command.negative.decisions CZ.response.transactions firecommand.transactions firecomand.positive.decisions firecommand.negative.decisions fire.response.transactions road.company.trasactions geotechnician.transactions Source: Agrež, own research (2015) 102

125 Principal Component 1 includes the following most influential attributes: community member positive decisions, community member negative decisions, distress call transactions. Principal Component 2 includes the following most influential attributes: civil protection command positive decisions, fire brigade command transactions, fire brigade command positive decisions, fire command brigade negative decisions. Principal Component 3 includes the following most influential attributes: distress call centre transactions, distress call centre positive decisions. Chi-squared test (Table 3.24) revealed the following: community member positive and negative decisions significantly influenced knowledge flow directed towards emergency response entities; civil protection command positive decisions significantly influenced knowledge flow directed towards a flood-endangered entity; fire brigade command positive and negative decisions significantly influenced knowledge flow in both directions. What is more, activities and decisions made by the distress call centre, civil protection responders, fire brigade responders, road company and geotechnical company did not significantly contribute to the knowledge flow within the response process framework, even though they assumed an important role in the process scenarios. The results provided an interesting insight in the structure of learning groups divisions within the response system. They revealed that community learning is not always a mutual process, and three levels of mutuality were identified: none, present and strong. The most influential attributes, considering the process scenarios division, are: community member positive decisions, community member negative decisions, distress call transactions, CP command positive decisions, fire brigade command transactions, fire brigade command positive decisions, fire brigade command negative decision, distress call transactions, distress call positive decisions. Even though these attributes contributed to the division of process scenarios, not all of them have a significant influence on the spread of knowledge. The most influential variables in terms of the knowledge spread are: community member positive decisions, community member negative decisions, CP command positive decisions, fire brigade command positive decisions, and fire brigade command negative decisions. A general overview of the results shows that community learning is not always a mutual process. Learning mutuality corresponds with the activities and decisions made by key entities in the process: endangered entity and responding decision makers. Consequently, H4 cannot 103

126 be fully confirmed, considering some detected scenarios in which learning mutuality was not detected at all. Therefore, H4 is confirmed only partially. Table 3.24: Chi-Square test Variable A Variable B Pearson Chi- Square Significance Community member transactions Number of red knowledge links 0.31 no Community member transactions Number of blue knowledge links 0 no Community member positive decisions Number of red knowledge links yes Community member positive decisions Number of blue knowledge links 0 no Community member negative decisions Number of red knowledge links yes Community member negative decisions Number of blue knowledge links 0 no Distress call centre transactions Number of blue knowledge links 0 no Distress call centre transactions Number of red knowledge links 0 no Distress call centre positive decisions Number of red knowledge links 0.99 no Distress call centre positive decisions Number of blue knowledge links no Distress call centre negative decisions Number of red knowledge links 0 no Distress call centre negative decisions Number of blue knowledge links 0 no CP command transactions Number of red knowledge links 0 no CP command transactions Number of blue knowledge links 0.28 no CP command positive decisions Number of red knowledge links 0 no CP command positive decisions Number of blue knowledge links yes CP command negative decisions Number of red knowledge links 0 no CP command negative decisions Number of blue knowledge links 0 no CP response transactions Number of red knowledge links 0.56 no CP response transactions Number of blue knowledge links no Fire brigade command transactions Number of red knowledge links 0.2 no Fire brigade command transactions Number of blue knowledge links 0.76 no Fire brigade command positive decisions Number of red knowledge links yes Fire brigade command positive decisions Number of blue knowledge links yes Fire brigade command negative decisions Number of red knowledge links yes Fire brigade command negative decisions Number of blue knowledge links yes Fire brigade responders transactions Number of red knowledge links 0.18 no Fire brigade responders transactions Number of blue knowledge links no Road company transactions Number of red knowledge links no Road company transactions Number of blue knowledge links no Geotechnical company Number of red knowledge links 0 no Geotechnical company Number of blue knowledge links no Source: Agrež, own research (2015) 104

127 3.6 Testing Hypothesis 5 The fifth hypothesis stated that knowledge-based process pattern recognition model can be used for ensuring public safety. I applied the previously developed research approach to the new, independent case in order to test whether it is possible to expand the methodology and insights from the previous research findings Background information The missing person investigation case used for the purposes of this research is based on events that took place in the period from May 23 29, The entities involved in the missing person investigation were the following: victim, victim s family, the bank, police patrol, police call centre, police station reception, family friend, human rights ombudsman, prosecution service, individuals, online community Search and Rescue (SAR) responders, and emergency call centre. I allocated the entities in groups and assigned them ID numbers, as seen in Table Due to privacy concerns, no personal information that could reveal the identity of individuals involved in the case will be revealed. The time scope of the incident was set at seven days, according to most important activities that took place within the identified time frame. Even though the roots of the incident go back into the past and the consequences could exist long into the future, they will not be included in the research scope due to their indirect connection with the topic presented. Missing person incident (MPI) activities that remain in the scope of the research can be divided into seven days. Day 1: victim cleared their internet history, temporary internet files, cookies and trash bin content with Ccleaner a software tool for PC optimization and cleaning. Day 2: victim left their home between 8 a.m. and 3 p.m. Victim was seen in public twice on that day, the first contact was 0.5 km from home and the second contact, at around 4 p.m., was 1.8 km from home, when the victim was heading approximately in the NNW direction. Day 3: victim had been identified at a cash machine surveillance camera recording, approximately 40 km from home. A new direction of the victim s movement was NW. Day 4: victim s family members asked for SAR responders to support them in the search. Day 5: victim had been seen 66 km from home. The heading direction was the same NW. The victim s family made contact with the human rights ombudsman and the prosecution services due to their dissatisfaction with police work. At around 10 p.m., the family received an from the victim, explaining that they were alive and expressing their dissatisfaction with life. Day 6: victim had been found by 105

128 family members approximately 80 km from home in a homeless shelter. Day 7: victim made contact with SAR responders and eventually met them to discuss the situation and decide how to close the case in a constructive way. Table 3.25: Identified entities Entity Entity ID Entity group Victim 1 / Family 7 / Bank 2 / Individuals 3 Community Online community 4 Community Human rights ombudsman 5 Public service Prosecution service 6 Public service Police patrol 10 Public service Police call centre 11 Public service Police station reception 12 Public service Emergency call centre 14 Public service SAR responders 13 / Family friend 8 Private support Individual members of the SAR team 9 Private support Source: Agrež, own research (2015) Data Primary data was collected directly on the field as a result of the possibility to observe the whole process from the early beginning on May 23, until the end on May 29. I was able to identify and note all entities that had a role in the process, and I tracked the timespan of the process and consequently defined its details on the level of individual activities. Not only was I able to link the activities conducted by the entities involved, I also had the possibility of following communication among different entities and of intercepting the information and knowledge interchange due to constant monitoring of the process. Based on the collected data, a process model was designed which served for further analysis, simulation and additional data, acquired through the dynamic process simulations. 106

129 3.6.3 Methodology As a research method, a case study contributes to a better understanding of a complex research object and emphasizes a detailed contextual analysis of a limited number of events or conditions and their relationships (Yin, 2003). According to Rahman et al. (2003), findings from case studies reveal a set of real activities at the particular moment of the research and are useful for building new theories, especially within the exploratory research. I decided to use the case study method due to a wide set of involved entities, an activity flow that is usually not present during the investigation process, and, consequently, a knowledge dynamic that was unique to this specific case. In order to collect the initial data based on a real-life missing person investigation process, I used the method of observation, as described by Mackellar (2013), which also allowed me to conduct the method of process tracing (Mahoney, 2010). I was in direct contact with most of the entities involved in the process, providing me with the ability to record their activities, attributes and interactions. I collected information about the marginal entities that were out of the scope of direct observation with semi-structured interviews, including entities that shared a connection with the marginal circle. Using a similar approach as that of Vanderploeg et al. (2012), I used the semi-structured interview also as a method of collecting information about the knowledge that entities used while performing activities in the process. In that respect, I used conversation, discussion, and questioning to get an insight into the investigated topic. TAD methodology was used for process modelling. It is a simple concept for describing the organization using several tables (Damij, 2000). Originally intended for information systems development and business process reengineering, it can be adopted for modelling a complex system. The first and second phases of TAD methodology include the framework on capturing and mapping the system functionality with the following tables: Entity, Activity, and Task. TAD methodology enables a creation of a model with any number of agent entities that can incorporate any number of activities. In my case, an entity is defined as a decisionmaking individual, a pair or a group of people modelled on the basis of real life events. Their actions are limited by activities and decisions that are part of the modelled process. Therefore, such a model precisely summarizes the process reality and maps it in a digital environment. I selected the TAD methodology due to its simple and intuitive graphic presentation of the modelled process diagram that provides an open frame for a further modular upgrade. 107

130 According to Kang et al. (2003), the method used for knowledge mapping and modelling is based on the integration of a workflow-based knowledge map, and the TAD activity table with additional integration of the knowledge table. I identified knowledge connected to the activities in the process and connected it through the process diagram. This approach enabled me to assign specific knowledge to specific activities and to distinguish between knowledge sources. I transformed a process-knowledge model into a dynamic simulation using the igrafx process simulation environment, similarly to Xia and Sun (2013). It is fully compatible with the TAD methodology and the knowledge map upgrade of TAD. The igrafx enables designing any number of entities that can differ by their hierarchical structure, resource range, process inclusion intensity, and complexes of their decision making logic. The main motivation for using igrafx was its ability to support such an approach with the integrated pallet of useful statistical tools that evaluated the relations among different variables within the process. Through a simulation report, I obtained the data based on transactions that affected entities in the process. I used the simulation data for further analysis of the investigation process. A similar approach is described also by Kiesling et al. (2012), who used agent-based modelling and simulation in the context of innovation diffusion. Several simulation scenarios provided data that I further analysed with descriptive statistical methods, employing the approach of Marshall and Jonker (2010), and further tested with the multivariate analysis of variance, similar to Patel et al. (2013), to determine the significance of key knowledge sources and build a model of knowledge in the analysed process Results I developed the model by first identifying the entities that participated in the process as knowledge sources, activity executors, or both. Furthermore, I mapped the activity flow with the TAD Activity table, excluding decision making. With the two-tier (activity and decision) process model, I had the ability to manipulate knowledge sources. Allowing the use of a specific type of knowledge consequently means that the knowledge sources or knowledge influenced entities will be able to execute a specific activity. Further, blocking the knowledge leads to an inability to execute the activity either due to a lack of know-what or know-how. The process begins with a single input and closes at three different possible ends, depending on the plot of the implemented scenario. A process 108

131 also begins with the attribute of the entity victim, set to state=missing. Among the three predicted endings, only one is preferred. This ending triggers a change of the attribute to state=found, the other two endings produce no change in the victim s attribute. All activities are lined up in time sequence as recorded in real situations. I extended the TAD methodology, shown in Table 3.28 (Appendix 9), in order to directly allocate knowledge to process entities Table 3.26: TAD extension, knowledge table Knowledge link Entity Knowledge Type Description Input Output/influence 2 Victim Cleaning the computer Knowhow Understanding importance of removing digital clues Solid digital clues Damaged digital clues/family 9 Family Victim's habits Knowwhat Understanding change in the routine Common behaviour Suspicion that something is not right 10 Family How to start Knowhow Understanding the need to define change in routine Undetected clues Detected clues Source: Agrež, own research (2015). A new descriptive Knowledge table was added, consisting of the following categories: Knowledge link: shows to which activity knowledge is linked within the process; Entity: shows what entity is the source of knowledge; Knowledge: provides the title of knowledge; Type: categorizes knowledge into know-what or know-how; Description: provides a brief description of knowledge; Input: provides the input necessary to trigger knowledge; Output: provides output in close connection with the activity where knowledge is deployed. Even though I constructed a new descriptive layer focused on knowledge, I lacked an intuitive presentation. In order to meet the theoretical requirements of knowledge assessment through knowledge mapping and demand a simplicity of practical use, I designed the map using TAD activity tables as the basic platform, supported by the newly generated Knowledge table. At this point, I adopted another TAD upgrade. I extended the activity table notation with the diamond sign that indicates knowledge, present in the execution of a single activity and dashed connector that indicates knowledge interaction (Figure 3.21). In order to execute a two-tier dynamic simulation, I translated the activity table into igrafx process diagram. I also used a Knowledge table to design the input generators that are sending inputs into the process, creating a flow of transactions, and simulating knowledge-based 109

132 decisions that allow or block the development of activity flow in the process. Moreover, I simulated the identified knowledge clusters through the integration in the predefined process model. For detecting the importance of present knowledge, I executed a simulation with 20 scenarios, following the multi-scenario analysis concept discussed by Simões and Marques (2011). Figure 3.21: TAD Activity table extended with knowledge mapping notation Source: Agrež, own research (2015) 110

133 In each scenario, I tested how the presence or absence of single identified knowledge influenced the activity flow and the final process output. Through the simulation s analytical report, I received the count of transactions that passed every single entity within the process simulation run. Based on the results of the simulations, I was able to determine a general indicator that revealed if a scenario ended with first/second (state=missing) or third ending (state=found). When the number of transactions for the entity victim exceeds 14, the state missing turns into the state found; in all other cases, however, the state remains unchanged and the process ends with one of the previous endings. Table 3.27: Knowledge simulation, transaction count Knowledge link ID Entity Transaction count 14 Emergency call centre SAR responders Human rights ombudsman Police patrol Police call centre Police station reception Family Family friend Individual members of SAR team Prosecution service Bank Individuals Online community Victim Victim state F M F F M M F M M F M M F F M M M M F F Source: Agrež, own research (2015) 111

134 Considering the defined indicator of success, I extracted 10 different kinds of knowledge with an important influence on the process output. When any kind of knowledge from the extracted group was blocked in the simulation, ending that changes the state=missing to the state=found was never reached. Table 3.29 shows the transaction count in all scenarios where all identified knowledge was evaluated. Figure 3.22 indicates the distribution of transactions that passed process entities in simulated scenarios. A clear distinction was seen among the three different groups of scenarios. The first group (red line) consists of scenarios where the upper quartile does not rise above one. Within these scenarios, the victim s attribute remains in the state=missing. Figure 3.22: Numerical distribution of process transactions Source: Agrež, own research (2015) The second group (blue line) includes scenarios where the upper quartile reaches the value seven, indicating the obvious higher level of transactions within the simulation, but according to their maximum that remains of value 14, these processes do not end with the output state=found. In the third group of scenarios with the upper quartile value at a minimum of 8, (green line) state=missing inevitably changes to state=found, making this group the only group of scenarios reaching the preferred outcome. An almost identical median value can be seen in the second and third groups, revealing that 50% of all entities encounter less than 2 112

135 transactions in scenarios where the victim s attribute remains state=missing and where it changes to state=found. A similar median value and a small value difference of the upper quartile between the second group and scenario with the lowest upper quartile of the third group reveals the probability of a high similarity of several scenarios, even though they produce different outputs. The correlation plot (Figure 3.23) confirms a high correlation between several scenarios, e.g.: scenario with blocked knowledge ID71 correlates with the scenario where knowledge ID76 was blocked, but they produce the opposite outputs. Such correlation leads to the conclusion that, from a holistic perspective, specific knowledge holds a crucial influence on the final output of the process. Figure 3.23: Numerical distribution of process transaction correlations between different simulation scenarios Source: Agrež, own research (2015) To distinguish specific knowledge from entities that serve as key knowledge sources, I selected scenarios where blocked knowledge triggered the output state=found and extracted the most important knowledge with key knowledge sources, as shown in Table

136 Table 3.28: Extracted important knowledge and key knowledge sources Entity ID , 13 8, 9 Knowledge source Family SAR responders Individuals/online community Individual members of SAR team/family friend Knowledge link 9, 13, 71 19, 20, 35 53, 14 54, 62, 42 Knowledge Victim's habits Proper reaction Understanding the importance Understanding the importance of continuing with the search Knowledge Try to get public support Proper approach of sharing information Understanding the importance of searching at specific places Knowledge Where and how to search Conclusions based on the information Willingness to share Understanding the importance of sharing information Source: Agrež, own research (2015) Table 3.29: Multivariate tests Multivariate Tests a Effect Value F Hypothesis df Error df Sig. Pillai's Trace b Intercept Wilks' Lambda b Hotelling's Trace b Roy's Largest Root b Pillai's Trace b Victim Wilks' Lambda b Hotelling's Trace b Roy's Largest Root b a. Design: Intercept + Victim, b. Exact statistic, c. The statistic is an upper bound on F that yields a lower bound on the significance level. Source: Agrež, own research (2015) For the next step of multivariate analysis, I transposed data from Table 3.31 in order to test which entities were the most significant during the transaction-based activity flow. Dependent 114

137 variables were defined as all entities present in the activity flow that determined the response to the activity flow created by the entity victim, which was used as a factor for the analysis. Results are shown in Table 3.32 and Table Table 3.30: Tests of between-subjects effects Tests of Between-Subjects Effects Source Dependent Variable Type III Sum of Squares df Mean Square F Sig. SAR responders Human rights ombudsman Police patrol Police call centre Police station reception Family Victim Family friend Individual members of SAR team Prosecution service Bank Individuals Online community Emergency call centre Source: Agrež, own research (2015) Multivariate analysis of variance yielded four highly significant entities: human rights ombudsman, prosecution service, family, and private/sar individuals. In comparing the significance with the TAD knowledge table, it becomes evident that the entities human rights ombudsman and prosecution service added no knowledge-based value to the process. Both entities represent identical behaviour during the process and appear in the process only with the activity flow that would consequently change state=missing to state=found. Conversely, 115

138 the entities family and private/sar individuals represent not just a significant value of the activity flow, but are also identified as key knowledge sources. Figure 3.24: Model of significant knowledge, and knowledge sources in the process Source: Agrež, own research (2015) Based on knowledge source extraction, TAD knowledge table and a TAD activity table with a knowledge map, supported by the multivariate analysis of variance, I built a model (Figure 3.27) representing significant knowledge interaction among key knowledge sources during the process. All four knowledge sources are necessary for the preferred ending of the simulated process, while the knowledge that these sources possess represents an essential element of the successful process output (the victim s attribute change). The model is composed of eleven important knowledge records, where knowledge of scenario Phase 1 enables the process to move on from an early phase, and knowledge of the second phase, together with the interaction between two significant entities, triggers a necessary activity flow direction that consequently creates the difference in the final process output. I was able to successfully apply knowledge-process relation based methodology to from the disaster management case study to the public safety case study. Even though the cases differ in many aspects, the conceptual model is rough enough that I can transfer it from one research field of interest to another with slight methodological adjustments due to specifics of the research case. Therefore, H5 can be fully confirmed. 116