-

Complementing i* with Game Trees - Getting to Win- Win in Industrial Collaboration

Vik Pant

vik.pant@mail.utoronto.ca 1

Eric Yu

eric.yu@utoronto.ca 0 1 0 Department of Computer Science, University of Toronto , Toronto , Canada 1 Faculty of Information, University of Toronto , Toronto , Canada

Interorganizational coopetition describes a relationship in which two or more organizations cooperate and compete simultaneously. Actors under coopetition cooperate to achieve collective objectives and compete to maximize their individual benefits. Such relationships are based on the logic of win-win strategies that necessitate decision-makers in coopeting organizations to develop relationships that yield favorable outcomes for each actor. This paper illustrates the introduction of a new actor in a coopetitive relationship as one of the pathways to a positive-sum outcome. It uses a strategic modeling approach that combines i* goal-modeling to explore strategic alternatives of actors with Game Tree decisionmodeling to evaluate the actions and responses of actors. This paper demonstrates the activation of one component in this guided approach of systematically searching for alternatives to generate a new win-win strategy. An interpretive adaptation of an industrial scenario that is drawn from practitioner and scholarly literatures is used to explain this approach. This illustration focuses on the Industrial Data Space which is a platform that helps organizations to overcome obstacles to data sharing in a coopetitive ecosystem.

Coopetition Win-Win Design Modeling

Coopetition refers to concomitant cooperation and competition among actors wherein actors “cooperate to grow the pie and compete to split it up” [ 1 ]. Actors under coopetition simultaneously manage interest structures that are partially congruent and partially divergent [ 2 ]. Partial congruence emerges from actors sharing in certain common objectives while partial divergence emanates from each actor’s pursuit of self-interest. Coopetition has become “increasingly popular in recent years” [ 3 ] and is widely observed in various domains including business, politics, and diplomacy [ 4 ]. Coopetition is predicated on the rationale of positive-sum outcomes through which all actors are better off by coopeting rather than by purely competing or solely cooperating. This aspect of coopetition requires decision-makers in coopeting organizations to develop and analyze win-win strategies. In this paper, we apply a synergistic approach that combines i* goal-modeling with Game Tree decision-modeling to generate and discriminate win-win strategies in a structured and systematic manner. We use an interpretive adaptation of an industrial scenario [ 9-11 ] from practitioner and scholarly literatures to explain this approach.

Win-Win Strategies and Positive-Sum Outcomes Coopetition research originated in the field of economics where researchers applied concepts from game theory to explain the motivations of coopeting actors [ 5 ]. According to game theory, three types of results are possible in strategic relationships between actors: positive-sum, zerosum, and negative-sum [ 6 ]. In positive-sum outcomes all actors are better off, in negative-sum outcomes all actors are worse off, and in zero-sum outcomes some actors are better off while other actors are worse off [ 6 ]. These outcomes are correlated to distinct types of strategies that are adopted by actors in coopetitive relationships: win-win, win-lose, and lose-lose. Win-win strategies lead to positive-sum outcomes, lose-lose strategies result in negative-sum outcomes, and win-lose strategies yield zero-sum outcomes. Rational and self-interest seeking actors are likely to seek positive-sum or zero-sum (i.e., only if they are better off) outcomes. Therefore, these actors will only implement win-win or win-lose (i.e., solely if they are advantaged) strategies voluntarily. However, win-lose strategies are unsustainable in coopetitive relationships because some actors (i.e., those that are disadvantaged) will be worse off as a result. These actors are likely to withdraw or abandon a win-lose relationship and therefore, win-win strategies are the only durable options for sustaining coopetitive relationships. 3

Modeling Win-Win Strategies using i* and Game Trees Pant and Yu [ 7 ] proposed a modeling approach for generating and discriminating win-win strategies using i* and Game Trees. They note that, “while game trees support the depiction of payoffs they do not explicitly codify the reasons for those payoffs” [ 7 ]. However, “even though the internal intentional structure of an actor cannot be expressed directly in Game Trees it can be represented via i* Strategic Rationale (SR) diagrams” [ 7 ]. In this approach, i* SR diagrams are used to represent and reason about internal intentional structures of actors while Game Trees are used to express and evaluate moves and countermoves of actors. Therefore, “Game Trees and actor modeling with i* can be used together to achieve a deeper understanding of the decision space as well as to secure a stronger decision rationale” [ 7 ].

In the modeling phase, the actors (concrete and abstract), goals, tasks, resources, and softgoals are denoted in the i* SR diagram while the sequence of their moves and countermoves is codified in the Game Tree. Figure 2 depicts related i* SR diagrams and Game Trees of As-Is and To-Be scenarios of a coopetitive relationship between two business partners in the pharmaceutical industry. Model elements in black color represent the As-Is scenario (with two alternatives) and model elements in blue color depict additional model elements in the To-Be scenario. The model elements in blue color are generated by following the process depicted in Figure 1.

Start Represent Stakeholders that are Concrete Actors

as Agents and

Abstract Actors

as Roles i* SR model showing Actors and Roles Game Tree showing player sequence

Represent Focal

Player as

First Mover Identify additional actors Identify goals for each actor Identify alternative tasks for achieving each goal Identify softgoals for each actor, with priorities Identify contributions from tasks to softgoals Identify

Dependencies among actors d n e g e L

Modeling Technique

i* SR model Game Tree showing Goals showing and Tasks move sequence

Represent Sequence

of Moves as Decisions

Phase Start/End Process Decision

i* SR model showing dependencies Evaluate goal satisfaction

by propagating

labels i* SR model showing complete structure

Compute

Payoffs for Decision Paths

Is there a win-win strategy?

Add/Remove some actor

Generate a change in some actor s goal Generate

additional alternatives for achieving goals of some actor

Generate a change in softgoals of some actor Generate a change in relationships among two actors

Yes End (1, -1) (-1, -1) (1, 1)

Game Tree A Actor Alternative Dependency Link Help Contribution Link Hurt Contribution Link Task Decomposition Link Satisficed Denied

Unknown

Branded Pharmaceutical Company (BPC) and Generic Pharmaceutical Compounder (GPC) are two actors that play the role of Pharmaceutical Business Partner (PBP)1. BPC develops and markets prescription medicines based on its research and development initiatives as well as intellectual property (IP) protections (not shown*). GPC manufactures ingredients that are used in BPC’s medicines and produces medicines for BPC that BPC sells to pharmacies and hospitals (not shown*). GPC also markets generic medicines that are analogous to the prescription medicines that are sold by BPC only if no organizations assert exclusivity on those formulations via their IP protections (not shown*).

GPC depends on Market Forecasts of BPC (shown) so that GPC can approximate the upcoming requirements of BPC (not shown*). This helps GPC to plan its production runs based on medicines that BPC might contract GPC to produce (not shown*). BPC depends on the Production Traces of GPC (shown) to verify that GPC is only manufacturing those quantities of ingredients of BPC’s high margin medicines that are ordered by BPC (not shown*). This helps BPC to verify that GPC is not manufacturing extra quantities of those ingredients to produce substitute medicines that GPC can sell by itself (not shown*). Dependencies among BPC and GPC are shown as softgoals because each is satisficed from the perspective of the depender. A problem that can occur in such strategic alliances is of ‘knowledge expropriation’ due to ‘learning races’ where each organization wishes to ‘learn faster’ than its partners [ 4 ]. Data sharing among partners is crucial for collaboration to be mutually beneficial. However, the possibility of information asymmetry among partners can motivate each organization to secure its own data. For example, access to GPC’s confidential production traces can endow BPC with bargaining leverage over GPC while access to BPC’s proprietary market forecasts can bequeath GPC with negotiating advantage over BPC (not shown*). Therefore, each actor is likely to demand more data from the other actor than it is willing to give to that other actor. The goal model of the role PBC shows that, in the As-Is scenario (represented in black color), BPC and GPC can demand partner data from each other. Upon receiving a demand for data, BPC/GPC can share its data directly with its partner by giving full access to own data or it can withhold its data from its partner by secreting its own data while hoarding its partner's data. In the evaluation phase, payoffs in the Game Tree are estimated by analyzing softgoal satisfaction in the i* SR diagram. The i* SR diagram shows that each of PBP's strategies impact its softgoals differently. Labels are placed above softgoals to depict their satisfaction or denial resulting from a certain strategy. Each task has a single checkmark above it at a specific position (i.e., first, second, or third) denoting a unique strategy corresponding with that task. Labels in the first, second, and third positions above a softgoal represent the impact of the strategy corresponding with that position on that softgoal. For example, Share data directly (first position) Denies Own data be secured but Satisfies Collaboration be mutually beneficial. The strategy of withholding data helps the softgoal of own data be secured but it hurts the softgoal of collaboration to be mutually beneficial. Conversely, the strategy of sharing data directly helps the softgoal of collaboration to be mutually beneficial but it hurts the softgoal of own data be secured. The sub-tasks of the withholding data strategy and the sharing data directly strategy impact the softgoal data asymmetry be reduced differently. In the withholding data strategy, secreting own data helps data asymmetry be reduced but hoarding partner data hurts data asymmetry be reduced. In the sharing data directly strategy, demanding partner data helps data asymmetry be reduced but giving full access to own data hurts data asymmetry be reduced. The payoffs in the Game Tree can be used to detect the presence of any positive-sum outcomes. In the As-Is scenario, there are no win-win strategies because neither the sharing data directly strategy nor the withholding data strategy allow the PBPs to satisfice all softgoals. This motivates their systematic search for new alternatives that lead to a positive-sum outcome. 1. Coopetition in the pharmaceutical industry is discussed in [ 8 ]. * This aspect of the relationship is not shown to keep the model simple.

In the exploration phase, a subject matter expert (SME) or domain specialist can pursue any of five non-deterministic lines of action incrementally and iteratively. As depicted in figure 1, they can add/remove some actor, generate additional alternatives for achieving goals of some actor, generate a change in relationships among two actors, generate a change in softgoals of some actor, or generate a change in some actor’s goal. Any of these actions can trigger other actions. For example, in the pursuit of a win-win strategy, an SME may choose to generate a new alternative that is an improvement over the best existing option (share data directly with payoff of 0,0). One possible improvement over this alternative is for the PBPs to share data via a mediating party if that party satisfies certain requirements. For the PBPs, sharing governance be transparent, is a crucial requirement and this can be satisfied by Industrial Data Space (IDS). The To-Be scenario is depicted in blue color. It shows IDS, which is a virtual data space that supports data-sharing commitments between partners under the purview of collaborative governance protocols [ 9-11 ]. PBPs can use IDS to co-develop data sharing protocols that protect their individual interests while advancing their mutual welfare. Each PBP must comply with data sharing protocols to which they commit in order to benefit from IDS’s capabilities that include tracking data exchanges, monitoring data usage, implementing data sharing protocols, and providing data sharing reports. This new alternative (share data via Industrial Data Space) in the To-Be scenario satisfices all the softgoals of PBP in the i* SR diagram and thus its payoff score in the Game Tree reflects a higher value than either of the options available in the As-I scenario. This new actor (IDS) triggers the creation of a new alternative for PBP and changes the interface of PBP by creating new dependencies between it and the new actor (IDS). 4

Conclusion We utilized an approach to systematically search for win-win strategies and generate new alternatives for organizations under coopetition. This integrative approach incrementally and iteratively elaborated and refined the i* SR diagram and its corresponding Game Tree. The resulting model explained the risk of knowledge expropriation in inter-partner learning arrangements if partners shared data directly. No win-win strategies were detected in the As-Is configuration because there existed the possibility for learning races where partners tried to learn faster than each other. In the To-Be scenario, a win-win strategy was generated by using the best-existing option as a starting point and reference. By sharing data via a mediating party, PBPs would benefit from collaborative information exchange without the risk of knowledge expropriation. 5

1. Brandenburger , A.M. , Nalebuff , B. J. ( 1995 ). The right game: Use game theory to shape strategy . Harvard Business Review 73.4 , 57 - 71 .

2. Castaldo , S. , Dagnino , G.B. ( 2009 ). Trust and coopetition: the strategic role of trust in interfirm coopetitive dynamics . In: Dagnino, G.B. , Rocco , E. (Eds.), Coopetition Strategy: Theory, Experiments and Cases. Routledge, New York, 74 - 100 .

3. Gnyawali , D.R. , Park , B. ( 2009 ). Co-opetition and technological innovation in small and medium-sized enterprises: A multilevel conceptual model . Journal of Small Business Management 47.3 , 308 - 330 .

4. Pant , V. , & Yu , E. ( 2018 ). Modeling Simultaneous Cooperation and Competition Among Enterprises . Business & Information Systems Engineering 60.1 , 39 - 54 .

5. Brandenburger , A. M. , Nalebuff , B. J. ( 1996 ). Co-opetition . New York: Doubleday.

6. Dixit , A. K. , & Nalebuff , B. ( 2008 ). The Art of Strategy: A Game Theorist's Guide to Success in Business & Life . WW Norton & Company.

7. Pant , V. , & Yu , E. ( 2017 ). Understanding strategic moves and reciprocity on Software Ecosystems: A strategic modeling approach . In: 9th International Workshop on Software Ecosystems (IWSECO 2017 ).

8. Baglieri , D. , Carfì , D. , & Dagnino , G. B. ( 2016 ). Asymmetric R&D Alliances in the Biopharmaceutical Industry: A Multi-Stage Coopetitive Approach . International Studies of Management & Organization 46.2- 3) , 179 - 201 .

9. Otto , B. , Auer , S. , Cirullies , J. , Jürjens , J. , Menz , N. , Schon , J., Wenzel, S. ( 2016 ). Industrial Data Space: Digital Sovereignty Over Data . München: Fraunhofer-Gesellschaft.

10. Jarke , M. ( 2017 ). Data Spaces: Combining Goal-Driven and Data-Driven Approaches in Community Decision and Negotiation Support . In: International Conference on Group Decision and Negotiation, 3 - 14 . Springer: Cham.

11. Quix , C. , Chakrabarti , A. , Kleff , S. , & Pullmann , J. ( 2017 ). Business Process Modelling for a Data Exchange Platform . In: Proceedings of the CAiSE'17 Forum, at the 29th International Conference on Advanced Information Systems Engineering (CAiSE 2017 ).