Mind Your Theory: Theory of Mind Goes Deeper Than Reasoning

TL;DR

This paper emphasizes the importance of understanding the depth of mentalizing in Theory of Mind tasks for LLMs, highlighting that current AI approaches overlook the initial step of determining when ToM should be invoked.

Contribution

It identifies a gap in current AI ToM research, advocating for evaluation methods that consider the dynamic decision of invoking ToM and its depth, inspired by cognitive science.

Findings

Current AI ToM work focuses mainly on inference, neglecting when to invoke ToM.

Existing benchmarks treat ToM as static logic problems, missing dynamic aspects.

Proposes new evaluation approaches inspired by cognitive tasks.

Abstract

Theory of Mind (ToM) capabilities in LLMs have recently become a central object of investigation. Cognitive science distinguishes between two steps required for ToM tasks: 1) determine whether to invoke ToM, which includes the appropriate Depth of Mentalizing (DoM), or level of recursion required to complete a task; and 2) applying the correct inference given the DoM. In this position paper, we first identify several lines of work in different communities in AI, including LLM benchmarking, ToM add-ons, ToM probing, and formal models for ToM. We argue that recent work in AI tends to focus exclusively on the second step which are typically framed as static logic problems. We conclude with suggestions for improved evaluation of ToM capabilities inspired by dynamic environments used in cognitive tasks.