Dynamic Epistemic Friction in Dialogue

TL;DR

This paper introduces the concept of dynamic epistemic friction in dialogue, modeling resistance to belief updates in collaborative AI, and demonstrates its predictive power in belief revision scenarios.

Contribution

It defines and formalizes dynamic epistemic friction within Dynamic Epistemic Logic, linking it to belief revision and demonstrating its application in dialogue analysis.

Findings

Model accurately predicts belief updates in dialogues

Friction measure correlates with belief resistance

Framework accommodates complex real-world dialogue scenarios

Abstract

Recent developments in aligning Large Language Models (LLMs) with human preferences have significantly enhanced their utility in human-AI collaborative scenarios. However, such approaches often neglect the critical role of "epistemic friction," or the inherent resistance encountered when updating beliefs in response to new, conflicting, or ambiguous information. In this paper, we define dynamic epistemic friction as the resistance to epistemic integration, characterized by the misalignment between an agent's current belief state and new propositions supported by external evidence. We position this within the framework of Dynamic Epistemic Logic (Van Benthem and Pacuit, 2011), where friction emerges as nontrivial belief-revision during the interaction. We then present analyses from a situated collaborative task that demonstrate how this model of epistemic friction can effectively predict belief updates in dialogues, and we subsequently discuss how the model of belief alignment as a measure of epistemic resistance or friction can naturally be made more sophisticated to accommodate the complexities of real-world dialogue scenarios.