A Study of Belief Revision Postulates in Multi-Agent Systems (Extended Version)

2026-06-12 · Source: cs.AI updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Multi-Agent Systems · Depth: Expert, extended

Summary

This study investigates belief revision in multi-agent systems, specifically addressing how agents' beliefs change after one agent acquires a new belief about a state property. Utilizing a single multi-agent Kripke model, the research generalizes the classical AGM belief revision postulates and the DP postulates for iterated revision to this multi-agent context. The authors introduce a generalized "full-meet" multi-agent belief revision operator, proving it satisfies all generalized AGM postulates but not DP2 or Independence (IN) for iterated revision. A more sophisticated "event model-based" revision operator is also presented, which satisfies all generalized AGM postulates and, under specific conditions, most DP postulates. A further refined event-based operator, `*rb`, is shown to satisfy all generalized MBR postulates. This work, an extended version of a paper accepted at KR 2026, provides a formal framework for evaluating dynamic epistemic reasoning.

Key takeaway

For AI Scientists and Research Scientists developing multi-agent systems, understanding the generalized AGM and DP postulates is crucial for designing robust belief revision mechanisms. Your choice of revision operator, such as the `*rb` event-based model, directly impacts how agents handle new information and maintain consistent beliefs, especially in complex, uncertain environments. Carefully evaluate operators against these postulates to ensure predictable and rational agent behavior.

Key insights

Generalizing classical belief revision postulates to multi-agent Kripke models provides a formal framework for dynamic epistemic reasoning.

Principles

• Multi-agent belief sets are represented by single, pointed Kripke structures.
• Classical AGM postulates can be generalized for multi-agent belief revision.
• Iterated revision postulates (DP) also require multi-agent generalization.

Method

The paper defines multi-agent belief revision (MBR) based on a single Kripke model. It generalizes AGM and DP postulates, then proposes a "full-meet" operator and two event model-based operators (`*ev`, `*rb`) for MBR.

In practice

• Evaluate multi-agent belief revision operators against generalized postulates.
• Implement `*rb` operator for robust multi-agent belief revision.
• Consider event models for dynamic epistemic reasoning in AI agents.

Topics

• Multi-Agent Systems
• Belief Revision
• Kripke Models
• Dynamic Epistemic Logic
• AGM Postulates
• Iterated Revision

Best for: AI Scientist, Research Scientist

Related on AIssential

• R-Mod: Minimal Structural Revision of S5 Epistemic Models — R-Mod minimally structurally revises S5 epistemic models to accommodate new knowledge, addressing limitations of belief revision.
• Belief Engine: Configurable and Inspectable Stance Dynamics in Multi-Agent LLM Deliberation — The Belief Engine provides auditable, parameter-controlled stance dynamics for LLM agents in deliberative simulations.
• An Abstract Worlds Semantic Framework for Belief Change Operators — Abstract Worlds Semantics provides a syntax-agnostic, set-theoretic framework for unifying diverse belief change models.
• Critic-Guided Heterogeneous Multi-Agent Reasoning for Reliable Mathematical Problem Solving — Critic-guided, heterogeneous multi-agent LLM systems significantly improve mathematical reasoning accuracy by iteratively correcting errors.
• Structured Abductive-Deductive-Inductive Reasoning for LLMs via Algebraic Invariants — A symbolic scaffold improves LLM logical reasoning by separating inference types and enforcing consistency via algebraic invariants.
• Text vs. K-Graphs: Why Your Multi-RAG System is Failing — LLMs exhibit severe biases, prioritizing information format over factual accuracy when resolving knowledge conflicts.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.