[Paper Review] Dispute Resolution Using Argumentation-Based Mediation

Tomas Trescak, Carles Sierra+2 more|arXiv (Cornell University)|Sep 15, 2014

Multi-Agent Systems and Negotiation25 references3 citations

TL;DR

This paper proposes an argumentation-based mediation framework that enables a trust-worthy computational mediator to resolve deadlocks in multi-agent negotiations by leveraging its own knowledge and resources to generate alternative solutions. Extending a logic-based BDI agent model, the mediator uses structured argumentation to propose actionable plans, enabling agents to achieve their goals even when initial negotiation fails due to knowledge or resource gaps.

ABSTRACT

Mediation is a process, in which both parties agree to resolve their dispute by negotiating over alternative solutions presented by a mediator. In order to construct such solutions, mediation brings more information and knowledge, and, if possible, resources to the negotiation table. The contribution of this paper is the automated mediation machinery which does that. It presents an argumentation-based mediation approach that extends the logic-based approach to argumentation-based negotiation involving BDI agents. The paper describes the mediation algorithm. For comparison it illustrates the method with a case study used in an earlier work. It demonstrates how the computational mediator can deal with realistic situations in which the negotiating agents would otherwise fail due to lack of knowledge and/or resources.

Motivation & Objective

• Address the limitation of existing argumentation-based negotiation systems where agents fail to resolve disputes due to insufficient knowledge or resources.
• Enable automated mediation in realistic negotiation scenarios where agents cannot independently find viable solutions.
• Design a computational mediator with access to domain knowledge and resources to propose alternative, negotiable solutions.
• Ensure the mediator's proposals are grounded in logical argumentation, allowing agents to reason about acceptability and feasibility.
• Demonstrate the mediator’s effectiveness in a case study involving goal-oriented agents with conflicting resource needs.

Proposed method

• Extend the BDI agent architecture from prior work to include explicit representation of resources and goals.
• Implement a mediation algorithm that identifies negotiation deadlocks and generates alternative solutions using the mediator’s knowledge and resource pool.
• Use formal argumentation frameworks to construct and exchange justifications for proposed solutions, with premises derived from domain rules and agent beliefs.
• Apply bridge rules (R.1–R.5) to translate mediated arguments into actionable requests (e.g., Ask, Tell) that agents can process and respond to.
• Integrate a generosity theory (G.3) and goal alignment (G.7) to enable agents to accept resource transfers when alternative plans are viable.
• Leverage logical inference (modus ponens, particularization) and belief propagation to validate the mediator’s proposed arguments before communication.

Experimental results

Research questions

• RQ1 How can a computational mediator resolve negotiation deadlocks when agents lack sufficient knowledge or resources to reach an agreement?
• RQ2 What formal mechanisms enable a mediator to generate and justify alternative solutions using its own knowledge and resources?
• RQ3 How can argumentation-based negotiation be extended to include a third-party mediator that actively contributes to solution generation?
• RQ4 In what way do logical argument structures and bridge rules facilitate the transition from abstract proposals to executable agent actions?
• RQ5 What conditions ensure agent acceptance of mediated proposals, especially when they involve resource trade-offs?

Key findings

• The mediator successfully resolved a deadlock in a home improvement negotiation case where agents could not achieve their goals due to missing resources and knowledge.
• By proposing alternative plans—such as using a screwdriver instead of a nail—the mediator enabled agent β to achieve its goal of hanging a mirror without requiring the nail.
• The argumentation framework allowed the mediator to generate logically valid justifications for its proposals, which were then converted into actionable requests via bridge rules.
• Agent α was able to achieve its goal of obtaining a nail only after agent β accepted a resource trade based on the mediator’s proposal, demonstrating the effectiveness of the solution.
• The use of formal logic and belief propagation ensured that all mediated arguments were consistent and verifiable, increasing trust and acceptance among agents.
• The system demonstrated that mediation is effective even when agents operate under incomplete information, provided the mediator has access to comprehensive domain knowledge and resources.