Abstract: We discuss a formal framework for modelling how agents extract reliable information from stories and communications in which speakers may strategically distort the truth. Building on Rubinstein-Glazer’s models of persuasion, we formalize Halpern’s Tiglath-Pileser principle, based on the idea that, while biased, reported claims must maintain a connection with the truth.
Our approach accounts for how Tiglath-Pileser relations are formed, revised, and strategically manipulated during a narration/communication, giving rise to higher-order reasoning about the reliability and intentions of information sources.
We show how this machinery can account for Halpern’s interpretation of the biblical account of King David. The framework integrates tools from logic and game theory, contributing to the understanding of dynamics of belief formation under strategic information transmission, and providing formal foundations for the theory of mind in agentic AI.

Chair: Prof. Ernst Wit