Understanding Epistemic Language with a Language-augmented Bayesian Theory of Mind
This addresses the challenge of modeling human-like understanding of belief-related language for cognitive science and AI, though it is incremental as it builds on existing Bayesian and language methods.
The paper tackled the problem of how people interpret epistemic language about others' beliefs by introducing LaBToM, a cognitive model that combines Bayesian theory-of-mind with language decoding, and it showed high correlation with human judgments across various expressions, outperforming models like GPT-4o and Gemini Pro.
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.