Reason to Play: Behavioral and Brain Alignment Between Frontier LRMs and Human Game Learners

2026-05-08

AI Summary

This study investigates whether modern artificial intelligence systems can rapidly acquire abstract knowledge in novel environments and flexibly guide behavior, akin to humans. By analyzing human participants’ behavior and concurrent fMRI data during gameplay in a novel video game, the authors systematically evaluate large reasoning models (LRMs) against reinforcement learning and Bayesian cognitive models in terms of strategic gameplay, behavioral fit, and neural activity prediction. Results demonstrate that LRMs significantly outperform baseline models—by an order of magnitude—in both behavioral alignment and whole-brain activity prediction, including cortical and subcortical regions. This advantage stems from their internal representations rather than planning mechanisms, providing the first evidence that LRMs exhibit strong alignment with human learning behavior and neural dynamics in complex, naturalistic tasks.

Abstract

Humans rapidly learn abstract knowledge when encountering novel environments and flexibly deploy this knowledge to guide efficient and intelligent action. Can modern AI systems learn and plan in a similar way? We study this question using a dataset of complex human gameplay with concurrent fMRI recordings, in which participants learn novel video games that require rule discovery, hypothesis revision, and multi-step planning. We jointly evaluate models by their ability to play the games, match human learning behavior, and predict brain activity during the same task, comparing a suite of frontier Large Reasoning Models (LRMs) against model-free and model-based deep reinforcement learning agents and a Bayesian theory-based agent. We find that frontier LRMs most closely match human behavioral patterns during game discovery and predict brain activity an order of magnitude better than both reinforcement learning alternatives across cortical and subcortical regions, with effects robust to permutation controls. Through targeted manipulations, we further show that brain alignment reflects the model's in-context representation of the game state rather than its downstream planning or reasoning. Our results establish LRMs as compelling computational accounts of human learning and decision making in complex, naturalistic environments. Project page with interactive replays: https://botcs.github.io/reason-to-play/

Problem

• human learning
• artificial intelligence
• game learning
• brain alignment
• abstract knowledge

Innovation

• Large Reasoning Models
• brain alignment
• human-like learning
• fMRI prediction
• in-context representation