Can Large Language Models Simulate Human Cognition Beyond Behavioral Imitation?

2026-03-29

AI Summary

This study addresses a critical gap in the evaluation of large language models (LLMs), which often rely on synthetic or aggregated data that cannot discern whether models genuinely simulate individual human cognition. To this end, the authors construct the first benchmark for cognitive simulation grounded in the longitudinal research trajectories of 217 AI researchers, using their published papers as externalized representations of cognitive processes. The benchmark evaluates models’ capacity for cognitive transfer across domains and under temporal distribution shifts, and introduces a multidimensional cognitive alignment metric to quantify individual-level consistency. Through systematic evaluation of mainstream LLMs and their enhancement strategies, the work empirically reveals both the current capabilities and fundamental limitations of existing models in emulating authentic human cognitive patterns.

Abstract

An essential problem in artificial intelligence is whether LLMs can simulate human cognition or merely imitate surface-level behaviors, while existing datasets suffer from either synthetic reasoning traces or population-level aggregation, failing to capture authentic individual cognitive patterns. We introduce a benchmark grounded in the longitudinal research trajectories of 217 researchers across diverse domains of artificial intelligence, where each author's scientific publications serve as an externalized representation of their cognitive processes. To distinguish whether LLMs transfer cognitive patterns or merely imitate behaviors, our benchmark deliberately employs a cross-domain, temporal-shift generalization setting. A multidimensional cognitive alignment metric is further proposed to assess individual-level cognitive consistency. Through systematic evaluation of state-of-the-art LLMs and various enhancement techniques, we provide a first-stage empirical study on the questions: (1) How well do current LLMs simulate human cognition? and (2) How far can existing techniques enhance these capabilities?

Problem

• large language models
• human cognition
• behavioral imitation
• cognitive patterns
• individual-level cognition

Innovation

• cognitive simulation
• longitudinal benchmark
• cross-domain generalization
• cognitive alignment
• individual-level modeling