Common Knowledge in Interaction Structures

We consider two simple variants of a framework for reasoning about knowledge amongst communicating groups of players. Our goal is to clarify the resulting epistemic issues. In particular, we investigate what is the impact of common knowledge of the underlying hypergraph connecting the players, and under what conditions common knowledge distributes over disjunction. We also obtain two versions of the classic result that common knowledge cannot be achieved in the absence of a simultaneous event (here a message sent to the whole group).

💡 Research Summary

The paper investigates epistemic reasoning in multi‑agent communication systems by formalising interaction structures as hypergraphs. A hypergraph’s hyper‑edges represent groups of agents that can exchange a message simultaneously. Two variants of the framework are studied.

In the first variant every agent has common knowledge of the underlying hypergraph. This means that all participants know, and know that they know, the exact topology of the communication network. Under this assumption the possible worlds for each agent are constrained in the same way, and the common‑knowledge operator C behaves exactly as in the standard S5 modal logic. The authors prove that C is reflexive, transitive and symmetric, and, crucially, that it distributes over disjunction: C(p ∨ q) ↔ Cp ∨ Cq. The proof (Theorem 2) shows that when the hypergraph is public, any disjunction that is common knowledge must already be common knowledge of one of its disjuncts. This result clarifies a long‑standing intuition that public network information enables agents to decompose shared beliefs cleanly.

The second variant drops the assumption of common knowledge about the hypergraph. Agents only observe the messages they receive; the structure of the communication channels remains hidden. Consequently, each agent’s set of epistemic alternatives may differ, and while C still satisfies the S5 axioms, the distributive law over disjunction fails in general. The paper supplies a concrete counter‑example (Theorem 4) where C(p ∨ q) holds but neither Cp nor Cq does, demonstrating that meta‑knowledge of the topology is essential for the distributive property.

Beyond the distributivity analysis, the authors revisit the classic impossibility theorem: common knowledge cannot be generated without a simultaneous event that reaches all agents. In the first variant they show (Theorem 5) that if there is no event in which every participant receives the same message at the same time, then no non‑trivial formula can become common knowledge, even though the hypergraph itself is public. In the second variant the impossibility is even stronger (Theorem 7): when the hypergraph is private, the absence of a simultaneous broadcast guarantees that Cϕ is false for every non‑tautological ϕ. The proofs adapt the standard “no‑simultaneity” argument to the hypergraph setting, carefully handling the additional combinatorial possibilities introduced by group messages.

The paper concludes with a discussion of practical implications. For designers of distributed protocols, making the communication topology public (or broadcasting it explicitly) and ensuring a round of simultaneous broadcast are sufficient conditions for building common‑knowledge‑based consensus mechanisms such as Byzantine agreement or atomic commit. Conversely, if the topology must remain hidden for security or privacy reasons, protocols that rely on common knowledge are fundamentally untenable; designers must resort to cryptographic proofs, partial agreements, or asynchronous consensus techniques.

Future research directions suggested include (i) dynamic hypergraphs where links appear or disappear during execution, (ii) probabilistic message loss and approximate common knowledge, and (iii) security analyses of how hidden hypergraph information can be inferred by adversaries. Overall, the paper provides a rigorous, dual‑perspective treatment of common knowledge in interaction structures, clarifying exactly when and why common knowledge distributes over disjunction and when it is impossible to achieve without a simultaneous event.