Reasoning aligns language models to human cognition

2026-02-09

AI Summary

This study investigates whether large language models (LLMs) align with human cognition in decision-making under uncertainty and how chain-of-thought (CoT) reasoning influences their belief updating and active sampling behaviors. To this end, we design an active probabilistic inference task that decouples evidence sampling from reasoning and introduce a near-optimal policy as a benchmark. We construct an interpretable mechanistic model comprising four latent variables—memory, strategy, selection bias, and occlusion awareness—to map both humans and LLMs into a shared low-dimensional cognitive space. Results show that CoT significantly enhances the models’ reasoning capabilities, yielding belief trajectories more closely aligned with human patterns, though it offers limited improvement in active sampling. The proposed mechanistic model effectively reproduces multi-agent behavioral characteristics, highlighting CoT’s pivotal role in achieving cognitive alignment.

Technology Category

• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning
• Reasoning under Uncertainty: Relational Probabilistic Models
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

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

Abstract

Do language models make decisions under uncertainty like humans do, and what role does chain-of-thought (CoT) reasoning play in the underlying decision process? We introduce an active probabilistic reasoning task that cleanly separates sampling (actively acquiring evidence) from inference (integrating evidence toward a decision). Benchmarking humans and a broad set of contemporary large language models against near-optimal reference policies reveals a consistent pattern: extended reasoning is the key determinant of strong performance, driving large gains in inference and producing belief trajectories that become strikingly human-like, while yielding only modest improvements in active sampling. To explain these differences, we fit a mechanistic model that captures systematic deviations from optimal behavior via four interpretable latent variables: memory, strategy, choice bias, and occlusion awareness. This model places humans and models in a shared low-dimensional cognitive space, reproduces behavioral signatures across agents, and shows how chain-of-thought shifts language models toward human-like regimes of evidence accumulation and belief-to-choice mapping, tightening alignment in inference while leaving a persistent gap in information acquisition.

Problem

• reasoning
• language models
• human cognition
• decision under uncertainty
• chain-of-thought

Innovation

• chain-of-thought reasoning
• active probabilistic reasoning
• cognitive alignment
• mechanistic modeling
• evidence accumulation