Argumentation-Based Negotiation

Unlike classical negotiation, ABN agents exchange supporting arguments and critiques alongside proposals. An agent doesn't just propose a price; it provides a justification (e.g., 'My offer of $X is fair because market data Y supports it'). This allows agents to reason about the acceptability of offers based on shared or attackable premises, moving beyond simple utility comparison.

ABN is grounded in computational models of argumentation, such as Dung's Abstract Argumentation Frameworks. Here, arguments are nodes in a graph, and edges represent attack relations (e.g., contradiction). The outcome of a negotiation is determined by evaluating which sets of arguments are collectively acceptable (e.g., 'admissible' or 'preferred' extensions) given these attacks. This provides a rigorous, logic-based semantics for determining when a proposal is justified.

A core objective of ABN is to influence and revise the preferences of other agents. By presenting persuasive arguments, an agent can change another agent's:

• Beliefs about the state of the world (epistemic arguments).
• Goals or desires (practical arguments).
• Valuations of outcomes. This leads to more informed agreements that may lie outside the original negotiation space, as agents' utility functions themselves can evolve during the dialogue.

ABN is particularly effective in domains with incomplete knowledge and uncertainty. Agents can use arguments to:

• Request missing information ('What is your evidence for that claim?').
• Challenge assumptions under uncertainty.
• Present defeasible rules that hold unless contradicted. This makes ABN robust for real-world scenarios where agents have partial, inconsistent, or evolving views of the environment and each other's constraints.

The negotiation follows a structured dialectical process—a formalized series of speech acts (e.g., assert, challenge, concede, retract). Agents engage in a dialogue game with rules governing permissible moves. This process explicitly surfaces the root of conflicts (e.g., conflicting beliefs about resource availability) and provides a protocol for resolving them through reasoned discourse, rather than mere compromise or breakdown.

ABN integrates with and complements other agent coordination patterns:

• Contract Net Protocol: ABN can be used within the bidding phase, where bids include justifications, and the manager's award decision is argument-based.
• BDI Architecture: Arguments directly target an agent's Beliefs, Desires, and Intentions.
• Distributed Constraint Optimization (DCOP): Arguments can justify constraint relaxations or preference changes to find feasible solutions.
• Electronic Institutions: ABN dialogues are often conducted within the normative rules of an electronic institution, which governs permissible argument types and outcomes.

A formal language with defined syntax, semantics, and pragmatics that enables autonomous agents to exchange information, requests, and arguments. FIPA ACL is the dominant standard, defining communicative acts like inform, request, propose, and refuse. It provides the grammatical foundation upon which argumentation protocols are built, ensuring messages are unambiguous and interpretable.

• Key Acts for Argumentation: assert, challenge, justify, retract.
• Semantics: Based on Speech Act Theory, where sending a message is performing an action (an illocutionary act).
• Message Structure: Includes fields for sender, receiver, performative, content, and conversation ID for protocol tracking.

A predefined, structured sequence of permissible message exchanges between agents to achieve a specific communicative goal, such as negotiation or inquiry. These protocols define the legal conversation states and transitions, often modeled as finite state machines or using Agent UML.

• Role in Argumentation: Provides the conversation framework that governs when agents can present offers, make arguments, critique, or concede.
• Examples: FIPA Contract Net Protocol (for task allocation), FIPA Iterated Contract Net, and custom protocols for multi-round argumentation.
• Formal Verification: Protocols can be analyzed to ensure properties like termination (the conversation always ends) and effectiveness (if a deal is possible, it will be found).

A prominent software model for intelligent agents where behavior is driven by its Beliefs (information about the world), Desires (potential goals), and Intentions (committed plans). This architecture is a common cognitive substrate for argumentation-capable agents.

• Argumentation's Target: Arguments aim to influence the beliefs and desires of other agents. A persuasive argument may cause an agent to adopt a new belief, altering its evaluation of options.
• Practical Reasoning: The BDI loop involves means-ends reasoning to form intentions, which is precisely what negotiation aims to align between parties.
• Implementations: The JACK and JADEX platforms are well-known BDI agent frameworks.

A formal framework for modeling problems where a set of agents must assign values to variables to optimize a global objective, subject to constraints, with decisions distributed among the agents. Argumentation can be used as a resolution mechanism within DCOP algorithms.

• Negotiation as DCOP: Each agent controls some variables with private constraints. The global goal is to find a joint assignment that maximizes overall utility.
• Argumentation's Role: Agents can exchange justifications (arguments) for their preferred assignments, revealing hidden utilities or constraints to find more optimal, mutually acceptable solutions.
• Algorithms: ADOPT, DPOP are classic DCOP algorithms; argumentation can enhance them by improving solution quality or speed of convergence.

Normative constructs that create obligations between agents, defining that a debtor agent is committed to a creditor agent to bring about a certain condition. They provide a formal foundation for trust and accountability in open systems.

• From Argument to Commitment: A successful argumentation-based negotiation often culminates in the creation of social commitments (e.g., "Agent A commits to delivering service X to Agent B for price Y").
• Dynamic State: Commitments have a lifecycle (active, violated, fulfilled, terminated). Arguments can be made about the status or viability of commitments.
• Regulative Framework: Commitments act as the enforceable outcome of persuasive dialogue, moving beyond mere communication to create binding social structures.

Computational frameworks that define the norms, rules, and structured interaction spaces (e.g., virtual rooms, auctions) governing the behavior of autonomous agents to ensure orderly and goal-directed societal interactions.

• Institutional Context: Argumentation-based negotiation typically occurs within the rules of an electronic institution, which defines what arguments are permissible, how they are evaluated, and the consequences of agreements.
• Components: Include a dialogical framework (protocols), a normative framework (rules and sanctions), and an ontological framework (shared vocabulary).
• Purpose: Reduces uncertainty and opportunism in open multi-agent systems by providing a predictable, rule-based environment for complex interactions like argumentative negotiation.