How FaR Are Large Language Models From Agents with Theory-of-Mind?

TL;DR

This paper introduces a new evaluation paradigm called Thinking for Doing (T4D) to test if large language models can use Theory-of-Mind in social scenarios, revealing their struggles and proposing a prompting framework to improve their reasoning and action selection.

Contribution

The paper presents T4D as a novel benchmark for assessing LLMs' ability to translate ToM inferences into actions and introduces the FaR prompting method to significantly enhance model performance.

Findings

GPT-4's performance improves from 50% to 71% with FaR.

LLMs excel at belief tracking but struggle with action guidance.

FaR outperforms Chain-of-Thought and Self-Ask methods.

Abstract

"Thinking is for Doing." Humans can infer other people's mental states from observations--an ability called Theory-of-Mind (ToM)--and subsequently act pragmatically on those inferences. Existing question answering benchmarks such as ToMi ask models questions to make inferences about beliefs of characters in a story, but do not test whether models can then use these inferences to guide their actions. We propose a new evaluation paradigm for large language models (LLMs): Thinking for Doing (T4D), which requires models to connect inferences about others' mental states to actions in social scenarios. Experiments on T4D demonstrate that LLMs such as GPT-4 and PaLM 2 seemingly excel at tracking characters' beliefs in stories, but they struggle to translate this capability into strategic action. Our analysis reveals the core challenge for LLMs lies in identifying the implicit inferences about mental states without being explicitly asked about as in ToMi, that lead to choosing the correct action in T4D. To bridge this gap, we introduce a zero-shot prompting framework, Foresee and Reflect (FaR), which provides a reasoning structure that encourages LLMs to anticipate future challenges and reason about potential actions. FaR boosts GPT-4's performance from 50% to 71% on T4D, outperforming other prompting methods such as Chain-of-Thought and Self-Ask. Moreover, FaR generalizes to diverse out-of-distribution story structures and scenarios that also require ToM inferences to choose an action, consistently outperforming other methods including few-shot in-context learning.