Language Models Represent Beliefs of Self and Others
This addresses the problem of understanding the mechanisms behind Theory of Mind capabilities in AI for researchers in cognitive science and AI, though it is incremental as it builds on existing knowledge of LLMs' ToM abilities.
The study discovered that neural activations in large language models can be linearly decoded to reveal internal representations of self and others' beliefs, and manipulating these representations significantly alters the models' Theory of Mind performance, with potential generalizability across diverse social reasoning tasks.
Understanding and attributing mental states, known as Theory of Mind (ToM), emerges as a fundamental capability for human social reasoning. While Large Language Models (LLMs) appear to possess certain ToM abilities, the mechanisms underlying these capabilities remain elusive. In this study, we discover that it is possible to linearly decode the belief status from the perspectives of various agents through neural activations of language models, indicating the existence of internal representations of self and others' beliefs. By manipulating these representations, we observe dramatic changes in the models' ToM performance, underscoring their pivotal role in the social reasoning process. Additionally, our findings extend to diverse social reasoning tasks that involve different causal inference patterns, suggesting the potential generalizability of these representations.