Understanding Mental States to Guide Social Influence in Multi-Person Group Dialogue

TL;DR

This paper introduces SocialMindChange, a benchmark for evaluating language models' ability to understand and influence mental states in multi-person social interactions, highlighting current models' limitations in dynamic social reasoning.

Contribution

The paper presents a new benchmark, SocialMindChange, for assessing how well language models can generate dialogue to change mental states in complex social scenarios.

Findings

Current LLMs perform 54.2% below humans on the benchmark.

Models struggle to maintain and modify mental states over long interactions.

The benchmark covers 1,200 social contexts with 90,000 questions.

Abstract

Existing dynamic Theory of Mind (ToM) benchmarks mostly place language models in a passive role: the model reads a sequence of connected scenarios and reports what people believe, feel, intend, and do as these states change. In real social interaction, ToM is also used for action: a speaker plans what to say in order to shift another person's mental-state trajectory toward a goal. We introduce SocialMindChange, a benchmark that moves from tracking minds to changing minds in social interaction. Each instance defines a social context with 4 characters and five connected scenes. The model plays one character and generates dialogue across the five scenes to reach the target while remaining consistent with the evolving states of all participants. SocialMindChange also includes selected higher-order states. Using a structured four-step framework, we construct 1,200 social contexts, covering 6000 scenarios and over 90,000 questions, each validated for realism and quality. Evaluations on ten state-of-the-art LLMs show that their average performance is 54.2% below human performance. This gap suggests that current LLMs still struggle to maintain and change mental-state representations across long, linked interactions.