Understanding Epistemic Language with a Language-augmented Bayesian Theory of Mind

TL;DR

This paper presents LaBToM, a Bayesian model that interprets epistemic language by integrating natural language processing with theory-of-mind reasoning, aligning closely with human judgments.

Contribution

Introduces LaBToM, a novel language-augmented Bayesian theory-of-mind model that interprets epistemic language through rational inference and language translation.

Findings

LaBToM correlates highly with human judgments across various epistemic expressions.

Outperforms multimodal LLMs and ablated models in predicting human belief attributions.

Effective in modeling graded plausibility of epistemic claims.

Abstract

How do people understand and evaluate claims about others' beliefs, even though these beliefs cannot be directly observed? In this paper, we introduce a cognitive model of epistemic language interpretation, grounded in Bayesian inferences about other agents' goals, beliefs, and intentions: a language-augmented Bayesian theory-of-mind (LaBToM). By translating natural language into an epistemic ``language-of-thought'' with grammar-constrained LLM decoding, then evaluating these translations against the inferences produced by inverting a generative model of rational action and perception, LaBToM captures graded plausibility judgments of epistemic claims. We validate our model in an experiment where participants watch an agent navigate a maze to find keys hidden in boxes needed to reach their goal, then rate sentences about the agent's beliefs. In contrast with multimodal LLMs (GPT-4o, Gemini Pro) and ablated models, our model correlates highly with human judgments for a wide range of expressions, including modal language, uncertainty expressions, knowledge claims, likelihood comparisons, and attributions of false belief.