Analyzing a Collaborative Modeling Game D. (Denis) Ssebuggwawo1 , S.J.B.A (Stijn) Hoppenbrouwers1 , and H.A (Erik) Proper1,2 1 Institute of Computing and Information Sciences, Radboud University Nijmegen Heyendaalseweg 135, 6525 AJ Nijmegen, The Netherlands, EU. D.Ssebuggwawo@science.ru.nl, stijnh@cs.ru.nl 2 Capgemini Nederland B.V., Papendorpseweg 100, 3528 BJ Utrecht, The Netherlands, EU. e.proper@acm.org Abstract. Analysing the modeling process within collaborative (group) modeling sessions is not a trivial task. In such environments, there are many things that influence the way the modeling process is carried along. These include the skills and expertise of the modelers, the communica- tion between them, the decision-making process, rules and goals driving the process etc. To study and support such a collaborative modeling pro- cess, we describe, in this work, a three-tier conceptual framework that uses the game-metaphorical approach. We present preliminary findings from a case study to illustrate the concepts in our framework. Keywords:Collaborative Modeling, Modeling Process, Quality of Model- ing, Modeling as a Game 1 Introduction In system development, including enterprise engineering, communication plays a vital role [9] and a number of stakeholders are usually brought aboard the system development ship with varying skills, expertise, and knowledge. This results in a heterogeneous group of stakeholders including, for example, project managers, (prospective) users who may act as domain experts etc. In such environments, participants engage in various types of conversations during the creation of agreed models. Such conversations involve negotiation, which results in accepts, rejects, modifications etc. see [2,6]. All this is done so that the different divergent po- sitions and viewpoints within the group can be reconciled, and agreement and consensus reached [9]. The work of Peter Rittgen ([6,7]) is closely related to our own, based on simi- lar principles, and therefore particularly relevant to this paper. His Collaborative Modeling Architecture(COMA) tool reflects a similar approach to collaborative system analysis and design. However, while he focuses on negotiation of models as such (which is indeed the core activity), he largely ignores other aspects (like language setting, planning, sub-model definition, etc.), or sets default choices for them. While we consider his approach a good start, we believe more differ- entiated and in-depth analysis of real modeling processes will contribute to a broader and deeper understanding of the modeling process. Proceedings of CAiSE Forum 2009 73 2 Proceedings of CAiSE 2009 Forum 2 Game Metaphorical Approach to Collaborative Modeling As discussed in [3], instantiated collaborative modeling sessions can be analyzed as if they are games. This approach is rooted in the observation that operational collaborative modeling is an interactive process that is “played out” within the boundaries of specific constraints (rules). Though our current analysis does not involve gaming as an overt activity, our analysis is based in the idea of viewing the modeling session that is studied as a game. Games, can be understood from many perspectives: systems, cognition, emo- tion (see, for example Jarvinen [4]); entertainment as well as utility (“games with a purpose” or “serious gaming”). Games are by definition rule/goal-oriented. Jarvinen developed a Game Design Theory (GDT) which, can be applied to method engineering [3]. Our work follows the same line of thought. Games Theory (which we do not directly employ) analyzes strategies for playing and winning games, whereas GDT describes design concepts and principles underlying good game design. In a modeling context, GDT is believed to contribute to good method design. As for the collaborative game aspect, [10] makes a clear distinction between competitive, cooperative and collaborative games. Competitive games force play- ers to identify strategies that are diametrically opposite. Cooperative games, in contrast to competitive games model situations in which players’ interests may be “neither completely opposed nor completely coincident”. They contain a set of enforceable rules that govern and direct the negotiation and bargaining of the players. Our work embraces this view and applies it to collaborative modeling by identifying a set of rules and goals governing and directing the modeling process, and studying the interactions in view of those rules. Thus, we view modeling as a game in which a set of rules and goals direct and govern the collaboration of the players (modelers). In this paper, we explain an operational conceptual framework related to the game metaphor, and illustrate the framework at the hand of data and results from a case collaborative modeling session. These being preliminary results we only report on results concerning interactions and few rules identified. 3 Conceptual Framework Our conceptual framework for analysis is based on previous theoretical work on the act of modeling [5], but pushes for operationalization of the theory in the form of qualitative analysis of (transcripts of) actual modeling sessions. In a col- laborative modeling session, modelers come together to perform some modeling task. They interact and communicate their ideas and opinions to other members. For them to reach consensus and agreement, they need to commit themselves to work as a team and abide by their collective knowledge, conventions and decisions (rules of their game). Their interactions, the rules and goals and the Proceedings of CAiSE Forum 2009 74 Proceedings of CAiSE 2009 Forum 3 Interactions lead to production of models, and generated (intermediate) models drive further interaction. Interactions Models Log <> <> A range of interactions over a period of time Some rules/goals of modeling apply to changes the rules of play and interactions are intermediary and end‐products and these guided and restricted by rules of play. Rules products may lead to new rules/goals. <> Fig. 1. A methodological approach for analyzing interactions, rules and models. models produced drive the modeling process at any given time, t. The interplay between them is given in Fig. 1. We view goals as a key type of rule (“goal rules”) [3]. The goals are rules setting states to strive for. The rules should ideally guarantee process and model quality, but they also reflect existing conventions for (inter)action in modeling and conversation. We distinguish two basic types of rules in collaborative mod- eling: rules set for the game, i.e., setting the game as such, and rules set in the game, i.e., by the players. 4 Collaborative Modeling Session Experiment The modeling session was organized and passively attended by two researchers. Three modelers (identified by M, D,R) participated in the actual session. They all had some experience in process modeling in a system development context, but were not expert modelers. The session (which took 18 minutes) was video recorded with good sound quality. The modelers were also given a digital writing pad, which was recorded alongside the video. See Fig. 2 for a snapshot of the recording. Transcription. A complete transcription was made of the recording and to effectively study the conversation patterns in the modeling session, we identified atomic interactions (i.e. disentangled them if they were wrapped up in complex sentences) and annotated and categorized them. The following information is covered by the coding of the transcript: • Interactions - with properties: time and interaction number, actor(player), topic/content, and speech act type. Table 1 gives an example of interaction coding, interpretation and the meta-data associated with its properties. Proceedings of CAiSE Forum 2009 75 4 Proceedings of CAiSE 2009 Forum Fig. 2. Snap-shot of the recording • Rules - with properties: time of activation, content and number of interac- tion it was proposed in, time of deactivation, content and number of interac- tion it was proposed in, type of rule. Interactions are identified by numbers. Table 1. Interaction coding, interpretation and meta-data Time/Int# Actor Speech Act - [Type] Topic 6:04 105 M Should we introduce a vendor, actor? [Question Set content 9] 6:08 106 D The material handler already functions as the Set content vendor. [Argue against 105] 5 Findings and Analysis The whole collaborative modeling session consisted of a total of 291 interactions and took 17.25 minutes or 1045 seconds. It showed three clearly distinguishable phases: I-Setting of the main approach: choosing the language and sub-division of work, II-Exploring and deciding which actors play a role in the first partial process model and III-Modeling the sub-process each with its own typical propor- tion of interactions types. A number of interaction topics and rules/goals were identified. These are shown and explained below. We used conversational analysis and Language-Action Perspective (LAP) [1] to identify the different speech acts in the modelers’ conversations. Proceedings of CAiSE Forum 2009 76 Proceedings of CAiSE 2009 Forum 5 Table 2. Number and type of speech acts within the phases Speech Act Type Phase I Phase II Phase III Total # % # % # % # % Propose/Answer 7 33 30 24 39 27 76 26 Counter propose 0 0 3 2 6 4 9 3 Question 7 33 25 20 16 11 48 16 Argue for 2 10 3 2 7 5 12 4 Argue against 1 5 9 7 7 5 17 6 Agree with/Accept 4 19 17 13 23 16 44 15 Disagree with/Reject 0 0 16 13 7 5 23 8 Non-verbal(graphical) 0 0 23 18 39 27 62 21 acts Total 21 126 144 291 Table 2 gives the distribution of the interactional speech acts over the three phases. In Table 3 we show the interaction topics as identified. The numbers and their corresponding percentages in the column total indicate the frequency use of the interaction topics. Table 3. Number and type of interaction topics Interaction Topic Phase GRM PLN CON CRT COL Total # % # % # % # % # % # % I 4 25 3 43 2 1 12 100 0 0 21 7 II 2 13 1 14 120 47 0 0 3 100 126 43 III 10 63 3 43 131 52 0 0 0 0 144 49 Total 16 7 253 12 3 291 GRM = Grammar, PLN = Planning, CON = Content, CRT = Creation, COL = Collaboration Categories of interactions identified are: Content Setting (which does not concern goal setting but fulfillment of goals) hence falls outside this paper’s scope, but deserves further study. Collaboration Setting is an interaction category not previ- ously proposed. It concerns how modelers are to collaborate with each other : what roles, hierarchy, responsibilities; how they organize themselves. Another “new” topic was found: Planning Setting, concerning options for temporal scheduling and strategies concerning the fulfillment of creation goals. We found nine rules, all goal setting rules. Three rules were explicitly set for the game: one creation rule, one grammar rule and one validation rule. Seven rules were set in the game: six of them concerned grammar goals, one concerned a creation goal in the game. Proceedings of CAiSE Forum 2009 77 6 Proceedings of CAiSE 2009 Forum 6 Conclusions and Further Research We have presented and illustrated a research approach aimed at analyzing the detailed process (act) of modeling. We presented a three-tier (Rules, Interac- tions and Models) conceptual framework. We analyzed an actual collaborative modeling session to illustrate the framework. Findings were also presented, to perform a partial validation of the Quality of Modeling (QoMo) [8] and COMA [7] approaches. We do not claim that our approach is definitive and static. There clearly is room for elaboration and improvement. We plan to carry on in this line of work in a recently started PhD project that this paper is also a product of. Our applied aim is to lay a foundation for the design of advanced, modeler-oriented support tools for collaborative modeling using a gaming approach to modeling. References 1. Goldkuhl, G.: Conversational Analysis as a Theoretical Foundation for Language Action Approaches? In H. Weigand, G. Goldkuhl, and A. de Moor, editors, Proceed- ings of the 8th International Working Conference on the Language Action Perspec- tive on Communication Modelling (LAP2003), Tilburg, The Netherlands. Springer (2003). 2. Hoppenbrouwers, S.J.B.A., Proper, H.A., Weide, T.v.d.: Formal Modelling as a Grounded Conversation. In M. Goldkuhl, G Lind and S. Haraldson, editors, Proceed- ings of the 10th International Working Conference on the Language Action Perspec- tive on Communication Modelling (LAP05)- Kiruna, Sweden, pp. 139-155, Linkpings Universitet and Hogskolan I Boras, Linkping, Sweden, EU, June (2005). 3. Hoppenbrouwers, S.J.B.A., van Bommel, P., Jarvinen, A.: Method Engineering as Game Design-An Emerging HCO Perspective on Methods and CASE Tools. In work- shop proceedings of EMMSAD08: Exploring Modeling Methods for Systems Analysis and Design affiliated to CAiSE08, Montpellier, France, May (2008). 4. Jarvinen, A.: Games without Frontiers, Theories and Methods for Game Studies and Design. PhD Thesis. University of Tampere, Finland (2007). 5. Proper, H.A., Hoppenbrouwers, S.J.B.A., van Bommel, P.: A Fundamental View on the Act of Modeling. In J. Kizza, J. Aisbett, A. Vince, and T. Wanyama, editors, Advances in Systems Modelling and ICT Applications, volume 2 of Special topics in computing and ICT research. Fountain Publishers, Kampala, Uganda, August (2006). 6. Rittgen, P.: Negotiating Models. In J. Krogstie, A. L. Opdahl, and G. Sindre, editors, CAiSE 2007, LNCS vol. 4495, pp. 561-573. Springer (2007). 7. Rittgen, P.: Collaborative Modelling Architecture (COMA). http://www.coma.nu/ COMA_Tool.pdf. [Accessed on: 08/02/2009]. 8. van Bommel, P., Hoppenbrouwers, S.J.B.A., Proper, H.A.: QoMo: A Modeling Pro- cess Quality Framework Based on Sequal. In H. A. Proper, T. Halpin, and J. Krogstie, eds., Proceedings of EMMSAD07, pp. 118-127. Tapir Academic Press, Trondheim, Norway (2007). 9. Veldhuijzen van Zanten, G., Hoppenbrouwers, S.J.B.A., Proper, H.A.: System De- velopment as a Rational Communicative Process. Journal of Systemics, Cybernetics and Informatics, vol. 2(4), pp. 47-51 (2004). 10. Zagal, J.P., Rick, J.: Collaborative Games:Lessons Learned from Board Games. Simulation & Gaming,Vol. 37 No. 1, 24-40. Sage Publications (2006). Proceedings of CAiSE Forum 2009 78