Mind Your Theory: Theory of Mind Goes Deeper Than Reasoning

2024-12-18

arXiv.org

AI Summary

Current LLM Theory of Mind (ToM) evaluation suffers from a fundamental bias—overemphasizing static logical reasoning steps while neglecting the critical prior decisions of *whether* to engage ToM and *what depth of mind* (DoM) is required. Method: The authors propose a novel, dynamic, environment-driven evaluation paradigm that systematically decouples ToM into two distinct phases: (1) *triggering and DoM selection*, and (2) *reasoning execution*. Their approach integrates ToM probes, modular enhancement mechanisms, formalized psychological models, and cognitive-science-inspired task design. Contribution/Results: Experiments reveal that mainstream benchmarks substantially underestimate LLMs’ deficiencies in depth-adaptive ToM decision-making. This work rectifies the prevailing static, logic-centric modeling of ToM and establishes a foundational theoretical framework—along with scalable assessment guidelines—for metacognitive, socially intelligent agents.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

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.

Problem

• Assessing Theory of Mind in LLMs
• Distinguishing mentalizing depth levels
• Proposing dynamic evaluation methods

Innovation

• Deep Theory of Mind
• Dynamic Mentalizing Depth
• Cognitive Task Environments