Search papers, labs, and topics across Lattice.
This paper investigates the conditions under which a shared causal backbone can be identified from private causal knowledge among agents, focusing on a two-agent common-effect scenario. The authors find that under typical assumptions, local causal marginals do not uniquely determine the backbone, leading to infinitely many joint intervention kernels that can produce identical private reports. They propose that effective identification requires agents to communicate causally identified response functions, emphasizing the communication aspect as a primary challenge before addressing policy composition.
Infinitely many causal structures can yield the same private reports, revealing the critical role of communication in identifying shared causal knowledge.
The Relativity of Causal Knowledge (RCK) explains how a network of agents with different structural causal models can exchange causal knowledge through a shared interventionally consistent abstraction, or backbone. We ask the prior identification question that this transport mechanism presupposes: when is that backbone determined by the agents' private causal knowledge? In the basic two-agent common-effect case, two private causes influence one shared outcome and each agent identifies only the single-cause causal marginal relevant to its own perspective. We show that, under standard compatibility, non-degeneracy, and local overlap assumptions, those local causal marginals do not identify a unique backbone. Infinitely many joint intervention kernels can induce exactly the same private reports while disagreeing on joint interventions. We then give a conditional recovery result. Additive separability removes the hidden interaction degree of freedom, but observational residual summaries remain insufficient. Identification becomes possible when agents communicate causally identified response functions. An education value-added example illustrates why this is first a communication problem, and only then a policy-composition problem.
