Varieties of Distributed Knowledge

2025-05-10

AI Summary

Distributed knowledge in multi-agent systems suffers from inconsistent definitions and semantic ambiguity across the literature. Method: We systematically survey and formally characterize twelve prominent definitions of distributed knowledge, constructing a unified logical modeling framework grounded in agents’ varying information-sharing capabilities. Using multi-agent modal logic and rigorous semantic analysis, we prove equivalence and entailment relations among these definitions. Contribution/Results: This work achieves the first fine-grained classification and comparability-based modeling of distributed knowledge, revealing its intrinsic logical hierarchy. We establish formal equivalence classes and a strict entailment graph, providing the first systematic taxonomic framework for distributed knowledge. This significantly advances the formal foundations of group knowledge in multi-agent logic, enabling precise reasoning about collective epistemic states and facilitating future theoretical and applied research in distributed epistemology and coordination.

Technology Category

Knowledge Representation and Reasoning: Description LogicsMultiagent Systems: Distributed Problem SolvingData Mining & Knowledge Management: Representing, Reasoning, and Using Provenance, Trust

Application Category

Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactionsGraph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphsWeb Mining and Content Analysis: Models for Web evolution

Abstract

Distributed knowledge is one of the better known group knowledge modalities. While its intuitive idea is relatively clear, there is ample room for interpretation of details. We investigate 12 definitions of distributed knowledge that differ from each other in the kinds of information sharing the agents can perform in order to achieve shared mutual knowledge of a proposition. We then show which kinds of distributed knowledge are equivalent, and which kinds imply each other, i.e., for any two variants $ au_1$ and $ au_2$ of distributed knowledge we show whether a proposition $phi$ being distributed knowledge under definition $ au_1$ implies that $phi$ is distributed knowledge under definition $ au_2$.

Problem

• Investigates 12 definitions of distributed knowledge
• Compares equivalence and implications between knowledge variants
• Explores agent information sharing for mutual knowledge

Innovation

• Investigates 12 distributed knowledge definitions
• Compares equivalence among knowledge variants
• Analyzes agent information sharing types