Beyond Common Knowledge: A Belief-Reasoning Framework for Non-Equilibrium Games and Its Graphical Representation (Extended Abstract)

Abstract

Classical Game Theory heavily relies on the “Common Knowledge” assumption, which mandates that all players know the game structure and others’ rationality ad infinitum. However, in asymmetric information scenarios, this assumption often collapses. This paper proposes a Belief-Reasoning Knowledge Model that decouples “States” from “Observable Features.” We introduce the Belief Forest, a graphical data structure to visualize and compress nested belief chains. We demonstrate the model’s efficacy through the “Hat Game” and a specialized Full-Belief Static Game analysis of the “Huarong Path” (Huarongdao) problem. We show that agents can achieve belief-driven strategies by assigning probability weights to belief trees, even in the absence of common knowledge.