Francisco Barbedo1,2, António Lucas Soares2,3

Arquivo Distrital do Porto , INESC Porto2 and Faculty of Engineering Univ. Porto3


This paper addresses the modelling of the social structure of collaborative networks and its relationship with inter-organizational business processes modelling and management. As the result of an action-research based process, an approach to the modelling of collaborative networks based on social actors' networks is proposed and its articulation with processes analysis and design is outlined. The collaborative network used in this study is composed of three organizations, both public and private, in the sector of port wine commercialization. We begin by presenting a short introduction to social actors' networks modelling and afterwards we describe the case study and some of the relevant findings.

"Networks are the fundamental stuff of which new organizations are and will be made" (Castells, 1996). This conclusion is more and more verified nowadays and for sure it will be in the future. The network is the fundamental unit of analysis (Castells, 1996) in economical, social, political and technological terms. In this research work we propose to recall some relevant findings in a well established area of social analysis to be applied to the analysis and design of Collaborative Networks (CN), including an important degree of computer mediated interaction. We will consider here CN as encompassing structural and process dimensions. The process dimension alone will not be addressed here as it is extensively analysed in the literature. Our interests are in the structural (social) dimension of CN, and in the relationship between structure and processes.

Social structure is related with regularities in relational patterns between concrete entities (Knoke and Kuklinsky, 1990). To fully understand the structure and dynamics CN, for example in a virtual organization context, we need theoretical and conceptual tools that help analysts and the network members themselves to manage

in the best way the processes undertaken by those networks. Structural modelling of social interaction has been used for a long time in social sciences disciplines. We believe that the impressive body of knowledge created in this are can be of particular importance to understand and manage better collaborative networks and virtual organizations. 2.1 Social actors' networks The analysis based on social actors, is based primarily in the assumption that a social structure does not organize itself in a random way but following certain patterns (Freeman, 2000). The social behaviour of actors, at any aggregation level (individual, group) is manifested through patterns of interactions (latent or evident) that are dependent on established connections between actors. A network model is then adopted to represent the social structure and its characterization can be obtained by the use of statistic and mathematical methods. 2.2 Organizational analysis perspectives SAN can be particularly useful in organizational (re)design, and reports of its effective use in that area are not new. For example, Tichy (1981) evaluates organizations using several network strategies after identify groups (“coalitions” and “cliques”) as well as their connections. It is possible then through SAN based methods to determine which groups are dominant, how they agree or disagree and, consequently if it is required to adopt cooperation strategies or conversely it is more adequate to establish negotiation scenarios. In general terms it is possible to use SAN to analyse organizational views, e.g. social, political, cultural, technological, at different levels - inter-organizational, organizational, group, individual (Tichy, 1981). Although a structural approach, SAN based methods enable to go well beyond formal, explicit structures: they can reveal the “informal” organizational chart (Mintzberg and Heyden, 1999; Molina, 2001). 2.3 SAN analysis in the context of inter-organizational processes As initially outlined in Soares and Sousa (2002), three analysis levels are possible with SAN: i) structural level tries to globally describe and analyse a given network, requiring complete data about the different actors and types of relationships. In the case of a complete supply chain, all the participating companies should be known, as well as all their client-supplier relations. Such analysis would for example be able to identify companies with common features (types of products, delivery performance, supply chain tier, etc.). ii) relational level concerns the description of the relationships between the social actors. A large number of attributes may be used to characterise these relationships (e.g. distance, accessibility) allowing an analysis that may be both qualitative and quantitative. Examples of this are the set-up and study of indicators for the logistic performance in operations involving two supply chain partners (such as delivery lead times, quality performance), the degree of trust between partners, the level of information exchange (such as production plans, informal data on production status).

Combining Social Structure and Process Analysis in Collaborative Networks iii) At the individual level, social actors are studied based on the relationships in which they participate (this characterisation should probably be complemented by other analysis methods). Examples are issues such as the importance of an actor (e.g., measured by the centrality of his position in terms of information exchange), or his potential for establishing relationships in the network. 2.4 Combining social structure with process analysis The role of a SAN in the analysis and design of inter-organizational processes is to reveal how the structure of the social relations between the actors performing the activities in the process can influence either positively or negatively the performance of the process. An illustrative example is described in section 4. One crucial aspect is the identification and characterisation of the social actors. The approach followed in this work was to select the social actors based on the participation and/or influence in the inter-organizational processes. This implies that the first step in data collection is the processes identification, which is a complex task on its own. We will not detail the process identification and mapping here as these methods are also extensively reported in the literature (see e.g. Eriksson and Penker, 2000).

This section describes the application of the concepts and methods described above in the analysis and design of a recordkeeping network, in the specific sector of Port Wine commercialization. 3.1 An action-research approach This research work followed an action-research approach. The main goal was to analyse a network recordkeeping organizations in order to reengineer their processes. This would involve, besides the business processes specification, the analysis and design of an informatics infrastructure for the network. It was then decided to depart from the following research questions: 1. Can a SAN based method be effective in the analysis of a social structure of a CN in the recordkeeping sector? 2. Can a SAN based method be combined with a business architecture modelling method to create a more comprehensive approach to interorganizational business processes (re)design? These two questions were the starting point of a spiral cycle of plan > act > observe > reflect aimed at solving the network problems from one side, and to create theoretical and practical knowledge from the other side. This research work occupied a person during 4 months in a time span of 12 months. 3.2 Organizational, social and legal context

Three organizations compose the network: ADP, the recordkeeping of Porto, IVP, the Port Wine Institute and EVP, a company commercializing Port Wine. IVP and EVP business is exclusively related with the port wine product, while ADP business is information and culture. EVP is private; the other two are public organizations. IVP and EVP are inherently linked by a legal framework, meaning that most of the social and functional relationships are imposed by a superior authority. This constrains the types of social relationships as well as the types of interorganizational processes to be performed. Nevertheless, the way how the processes are executed is prone to some variability. It is on this issue that the SAN analysis will bring some input regarding the optimization of those processes. 3.3 Actors selection and aggregation level The inclusion of actors in the network was made according to a nomothetic principle i.e. according to a set of criteria established a priori by the analyst. These criteria were 1. to include every actor participating in an inter-organizational process and 2. to include those actors that are not participating but can influence its execution by some means. The selected level for the aggregation of actors was the functional unit. This is a compromise between departmental units which would be too coarse for the required analysis and the individual role that would cause an explosion of actors and would add unnecessary complexity to the models. The methods used to collect data were semi-structured interviews, observation, document analysis and group discussions. The last method was fundamental to involve the relevant people in the analysis process aiming at incorporating a more interpretative dimension to the analysis. Summary of the results of the SAN analysis After the data collection phase, the next step was to select the measures where to base the analysis of the network, in this specific context. Based on reported experiences (Ajuha and Carley, 1998; Hagen et al., 1998), we choose the following measures: • General measures for characterisation of the network as a whole cohesion, density, transitivity; • Individual measures that take as a referential an individual actor and the connecting categories with the rest of the set - adjacency, geodesic, connectivity and maximum flow; • Centrality measures were used to determine the core actors in the network, the ones crucial for the development and support of the inter-organizational network; • Subgroup identification measures for a nClique analysis. A diagram of the actors and relationships between them is represented in Figure 1. Furthermore, darker actors have an higher weight and the relationships types are transaction and control.

Combining Social Structure and Process Analysis in Collaborative Networks

Figure 1 - Diagram of actors and relationships1 The global analysis enables the identification of the actors with higher weight (FISC, ENOL, COM, SAQ; PCIVP) and actors that are relatively isolated (GINFO e APROV). The network has a unique component (it is a completely connected subgraph). The individual analysis provides results on adjacency, number of geodesics, maximum flow and connectivity, centrality, proximity and intermediation. As an example, it was found that the actor FISC is the more central in the network. This confirms the empirical and perceptual data and it is very simply explained by the relevance of supervisory and control activities that still are the main responsibilities derived from the IVP policy. In the subgroup analysis (nCliques), the goal is to identify sets of actors that have a high cohesion, that act in a reciprocal way, that share the same social values. This

Actors from IVP are inside a square, EVP inside a lozenge and ADP inside an ellipse alone. Legend: 1 PCIVP - planning and control IVP; 2 GCX - accountancy; 3 FISC - supervision; 4 SAQ quality audit; 5 - ATQ technical evaluation of quality; 6 - GPC cultural assets management; 7 - GINFO information management, analysis and forecast; 8 - JCP consultive committee of tasters; 9 - PCADP planning and control ADP; 10 - GINC incorporations management; 11 - GAT technical support; 12 PROD Production; 13 - MARKT Marketing; 14 - COM Commercialization; 15 - APROV Procurement; 16 - ENOL Enology; 17 - PCIVP planning and control EVP.

measure points to actors that are closely aggregated through the execution of functions or tasks, or through informal relationships, even if included in processes or institutional activities. Figure 2 shows a graphical representation of the 8 cliques found in this network. The inclusion of certain types of actors in a clique enables to infer the existence of a degree of cohesion based both on the specialization of tasks and in the co-participation in inter-organizational processes. For example, cliques K1 and K2 include actors clearly specialized in technological activities (ATQ, ENOL, JCP), side by side with actors playing roles of coordination and administrative support to the core set of specialized tasks. Two general conclusions should be highlighted: 1. There is a discrepancy between the actors identified as more influent and central in the network e those that have more power as concluded through the centrality of intermediation measure. This can be explained by the fact that centrality of hierarchy within a network is better analysed by through the capacity o influence than by the degree of active participation. 2. High centrality and weight of an actor corresponds to a frequent Figure 2 - nCliques and actors participation in inter-organizational processes. This means that the information infra-structure should reflect this. In the same way, the distribution pattern of cliques by specialization areas can be a starting point for the definition of the information system supporting the inter-organizational processes, e.g. workflow patterns. Example: Process 05 - Authorization of External Trading We give now an example of the link between SAN analysis and process analysis. The activities in the process Authorization of External Trading are performed by the actors COM (EVP) and FISC (IVP). From the SAN of Figure 1 (observe the actors inside the dotted ellipse) we derive that the relationship is multiplex because it encompasses connections where control is exerted unilaterally upon one actor, but also transactional interactions where resources are exchanged across the connection. Centrality measures point these two actors as occupying central positions in the network (this can be concluded simply from the diagram of figure 1). We can conclude that they play an important role in the network. From the process models (not included here for the sake of space) a set of improvements were proposed. These proposals were considered technologically and organizationally feasible. The next step, enabled by the SAN results, was to analyse the constraints to the implementation of the improvements imposed by the social structure. Summarising, it was concluded that it was needed a substantial change in

Combining Social Structure and Process Analysis in Collaborative Networks the type of relationships between FISC and COM namely to reduce the intensity of control relationship. Only this way the technological based improvement - electronic authorization - could be implemented.

This paper reports a brief overview of an extensive research work carried out within the described network context. Most of the findings until now lie mainly in the social structure analysis. The connection between social structure models and process models was already approached but our conclusion is that we did not achieve yet a logical and practical articulation of both modelling methods to be considered as a potential analysis and design method of inter-organizational processes executed by collaborative networks. We will continue the research this issue by further analysing the collected data and by observing other CN. Another important issue insufficiently addressed so far is the inclusion of information technology and systems in the social structure and process models. An outline of a possible approach was already pointed out by Soares et al. (2001) for systems based on software agents architectures. In this particular case, a new and challenging requirements analysis and specification approach will be possible. Some interaction features and other attributes of a community (or society) of software agents can also be modelled by SAN, which leads to a coherent and uniform analysis method, opening some analysis possibilities not normally considered in the field of agent-oriented software engineering. Acknowledgments The work presented in this paper was partially developed within the European Project THINKcreative – Thinking network of experts on emerging smart organizations (IST-2000-29478). The authors would like to thank the European Commission for funding, and their project partners for suggestions and comments.

AHUJA, K., Manju; CARLEY, M., Kathleen – Network Structure in Virtual Organizations [em linha]. Journal of Computer Mediated Communication, 3(4) (1998) [referência de 2002-08-11]. Internet CASTELLS, Manuel, 1996, The information age : economy, society and culture. 3 vol.Blackwell,cop. Vol. 1 : The rise of the network society.- XVII, 556 p. Vol. 2 : The power of identity.- vol.2: XV, 461 p. Vol. 3 : End of millennium.- vol.3: XIV, 441 p.. ISBN 0-631-21594-8. ERIKSSON and PENKER, 2000, UML Business Modelling: Business Patterns. Wiley. FREEMAN, C. Linton – "Social Network Analysis: Definition and History" [on-line] In KAZDAN, A. E. ed. Encyclopedia of Psychology. New York: Oxford University Press, 2000, Vol. 6, pp. 350-351. also available at < l >. HAGEN, Guy; KILLINGER, Dennis, K.; STREETER, Richard B. – An Analysis of Communication Networks Among Tampa Bay Economic Development Organizations. Connections 20(2) (1997) disponível na internet em < >

KNOKE, David; KUKLINSKY, James, H. – Network Analysis. 5ª ed. Newbury Park: SAGE. Series:University Papers on Quantitative Applications in the Social Sciences (28). 1990. ISBN 080391914X MINTZBERG, Henry; HEYDEN, Ludo van der – Organigraphs: Drawing how Companies Really Work. Harvard Business Review, Setembro-Outubro (1999). MOLINA, José Luis – The Informal Organizational Chart in Organizations: An Approach from the Social Network Analysis. [on line] Connections 24(1) 2001. Also available at SOARES, António L., TOSCANO, C., SOUSA, Jorge P., 2001, A Social Actors Network Approach for the Design of Networked and Virtual Enterprises, in Kovacs, Bertok and Haiddeger (Eds.) Digital Enterprise Challenges: Life-Cycle Approach to Management and Production, pp.337-348. Kluwer, Netherlands. SOARES, António L., SOUSA, Jorge P., 2002, Multiple Perspective Configuration of Virtual Enterprises Using Social Actors Networks, in Luis M. Camarinha-Matos (Ed.) Collaborative Business Ecosystems and Virtual Enterprises, pp. , Kluwer Academic Publishers, Boston. TICHY, Noel, M. - Networks in Organizations. In NYSTROM, Paul, C.; STARBUCK; HAYNES, William - Handbook of Organizational Design. Vol.2: Remodeling Organizations and their Environments. New York: Oxford University Press,1981. ISBN 0198272421