Computational Foundations for Strategic Coopetition: Formalizing Trust and Reputation Dynamics

AI summary

Overview

• Research area: Multi-agent systems and conceptual modeling — specifically the intersection of computational trust models, game theory for strategic coopetition, and goal/dependency modeling languages (i*).
• Technical level: Advanced. The work assumes familiarity with game-theoretic equilibria, trust/reputation algorithms in multi-agent systems, and conceptual modeling notations for actor dependencies.
• Scope: The paper (a technical report) develops a two-layer computational trust model with asymmetric updating, plus a translation framework from i* dependency networks, and validates it through large parameter sweeps and a longitudinal alliance case study.

What This Paper Is About

Many real-world systems involve stakeholders who must cooperate and compete at the same time — coopetition — and their trust in one another shifts as they observe each other's behavior over repeated interactions. Two research communities cover half of this problem each: conceptual modeling languages such as i* can describe trust relationships qualitatively but cannot compute how trust changes, while multi-agent trust models compute updates algorithmically but are not grounded in models that capture actors' mixed strategic motives. The paper's goal is to bridge that gap by formalizing trust evolution on top of a game-theoretic foundation for strategic coopetition, and by giving practitioners a defined way to move from an i* dependency network to a running computational trust model.

Key Contributions

1. A computational trust model extending game-theoretic coopetition foundations. Trust is formalized as a two-layer system rather than a single scalar: an immediate trust layer that responds to current observed behavior, and a reputation layer that accumulates violation history.

2. Asymmetric trust updating dynamics. The model specifies that cooperation raises trust gradually while violations degrade it sharply. The abstract identifies this asymmetry as the mechanism that produces hysteresis effects and "trust ceilings" that limit how far a damaged relationship can recover.

3. A structured translation framework from i to computation.* The authors define a procedure that lets practitioners instantiate the computational trust model directly from i* dependency networks that encode actors' mixed motives, connecting qualitative strategic modeling to executable trust dynamics.

4. Dual validation — synthetic and empirical. Large-scale parameter exploration is combined with a documented real-world alliance case study to test whether the modeled dynamics match reality.

Main Findings

• Robust emergence of negative asymmetry. Across 78,125 parameter configurations, the authors report robust emergence of negativity bias, hysteresis effects, and cumulative damage amplification — i.e., these behaviors arise from the model's structure across a wide parameter range rather than from a narrow tuning.

• Two-layer trust structure drives the dynamics. Separating immediate trust (current behavior) from reputation (violation history) is the design choice that supports both gradual trust-building and sharp erosion.

• Trust is path-dependent. Hysteresis and trust ceilings mean that a relationship's history constrains its future: recovery after violations does not simply retrace the path by which trust was lost.

• The translation framework is usable. The abstract presents the i* → computational model translation as a structured, practitioner-oriented procedure, though it does not give step-by-step detail in the abstract itself.

• Case-study reproduction of trust evolution. Applied to the Renault-Nissan Alliance from 1999 to 2025, the model achieves 49/60 validation points (81.7%), reproducing documented trust evolution across five distinct relationship phases, including crisis and recovery periods.

• Prior result inherited from companion work. The abstract notes that companion work achieved 58/60 validation (96.7%) for logarithmic specifications; that figure belongs to the companion work, not to the results reported here.

Methodology in Plain English

The authors take an existing game-theoretic account of strategic coopetition and extend it with a trust mechanism that operates on two levels at once. One level tracks how an actor feels about a partner's current behavior — this moves in small increments when things go well. The other level keeps a running record of past violations, which is what makes trust drop suddenly when something goes wrong. Because the two levels move at different speeds and in different directions, relationships become path-dependent: damage is faster than repair, and there is a ceiling on how much trust can be rebuilt. To make this usable rather than purely theoretical, the authors define a translation procedure that converts an i* dependency network — a diagram-style model of who depends on whom and for what, including actors with mixed motives — into an instantiated computational trust model. They then test the model in two ways: by sweeping a very large space of parameter settings to see whether the signature behaviors appear consistently, and by comparing the model's output against a real, long-running corporate alliance with a well-documented history of cooperation, crisis, and recovery.

Why This Matters

Impact on research. The paper argues for closing a long-standing gap between qualitative conceptual modeling and algorithmic trust computation. If trust dynamics can be derived from dependency models that already capture strategic mixed motives, then trust research and requirements/enterprise modeling research stop being separate literatures and can inform each other. The reported emergence of negativity bias, hysteresis, and cumulative damage across a very large parameter space suggests these are structural properties of the formalism rather than artifacts of particular settings.

Real-world applications. The abstract does not enumerate application domains; the case study is a corporate strategic alliance. Domains that share the described structure — repeated interaction, simultaneous cooperation and competition, observable behavior — would include, by way of illustration:

• Strategic alliances and joint ventures between firms that both collaborate on some activities and compete on others.
• Supply-chain and outsourcing relationships, where suppliers may also be rivals and where violation history shapes contract terms and monitoring.
• Platform ecosystems, where complementors and platform owners depend on each other while competing for value capture.
• Standards consortia, multi-party infrastructure projects, or other long-horizon partnerships where trust must survive periodic disputes.

Industry relevance. The translation framework is aimed explicitly at practitioners: it offers a route from a strategic dependency diagram that stakeholders can already draw and argue about, to a model that predicts how trust will evolve under different behaviors. That supports decisions about how much to invest in relationship repair, how much weight to give a single breach, and why recovery timelines are asymmetric with damage timelines.

Future Directions

• Extending beyond the current trust layers. The abstract describes exactly two layers (immediate trust and reputation); whether additional layers — for example, competence versus integrity trust, or third-party reputation propagation — improve fidelity is left open.
• Broadening empirical validation. The Renault-Nissan Alliance is a single, long-horizon case. The abstract reports 49/60 validation points but does not describe validation against other alliances, shorter-horizon relationships, or non-corporate settings.
• Formalizing the translation framework further. The abstract presents the i* → computational translation as a framework, but does not state its completeness, its handling of complex dependency structures, or tooling support.
• Connecting parameters to observable evidence. The abstract mentions 78,125 parameter configurations but does not say how a practitioner should estimate a given relationship's parameters from real behavioral data, which is the practical bottleneck for adoption.

Target Audience

Researchers in multi-agent systems, computational trust and reputation, and conceptual/requirements modeling — particularly those working on i* and related goal- and actor-oriented notations — will find the core contribution relevant. It is also aimed at practitioners doing enterprise or alliance modeling who want to move from qualitative dependency diagrams to something computable, and at organizational and strategy researchers interested in formal accounts of trust dynamics in coopetitive relationships. A game-theoretic and modeling background is assumed; readers without one will find the formalism the main barrier.

Authors’ abstract

Modern socio-technical systems increasingly involve multi-stakeholder environments where actors simultaneously cooperate and compete. These coopetitive relationships exhibit dynamic trust evolution based on observed behavior over repeated interactions. While conceptual modeling languages like i* represent trust relationships qualitatively, they lack computational mechanisms for analyzing how trust changes with behavioral evidence. Conversely, computational trust models from multi-agent systems provide algorithmic updating but lack grounding in conceptual models that capture strategic dependencies covering mixed motives of actors. This technical report bridges this gap by developing a computational trust model that extends game-theoretic foundations for strategic coopetition with dynamic trust evolution. Building on companion work that achieved 58/60 validation (96.7%) for logarithmic specifications, we introduce trust as a two-layer system with immediate trust responding to current behavior and reputation tracking violation history. Trust evolves through asymmetric updating where cooperation builds trust gradually while violations erode it sharply, creating hysteresis effects and trust ceilings that constrain relationship recovery. We develop a structured translation framework enabling practitioners to instantiate computational trust models from i* dependency networks encompassing mixed motives of actors. Comprehensive experimental validation across 78,125 parameter configurations establishes robust emergence of negativity bias, hysteresis effects, and cumulative damage amplification. Empirical validation using the Renault-Nissan Alliance case study (1999-2025) achieves 49/60 validation points (81.7%), successfully reproducing documented trust evolution across five distinct relationship phases including crisis and recovery periods.