Centaur: a foundation model of human cognition

2024-10-26

arXiv.org

AI Summary

This study aims to develop the first foundational cognitive model capable of generalizing across human psychological behaviors, thereby advancing the computational formalization of a unified cognitive theory. Method: We fine-tune large language models on Psych-101—a massive-scale natural-language dataset of psychological experiments (10+ million trials across 160 paradigms)—and introduce two key innovations: natural-language task formalization and neural representation alignment. Contribution/Results: Our model achieves, for the first time, zero-shot behavioral prediction across distinct experiments, tasks, and cognitive domains. It significantly outperforms classical cognitive models in held-out participant prediction accuracy. Crucially, its learned behavioral representations exhibit high fidelity to human neural activity measured via fMRI and MEG. As the first scalable, interpretable, and empirically falsifiable computational framework for cognitive science, it bridges symbolic cognitive modeling with neural data, enabling rigorous hypothesis testing and theory-driven AI development.

Technology Category

• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Natural Language Processing: (Large) Language Models

Application Category

• User Modeling, Personalization and Recommendation: Psychology-informed user models and recommender systems
• 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

Abstract

Establishing a unified theory of cognition has been a major goal of psychology. While there have been previous attempts to instantiate such theories by building computational models, we currently do not have one model that captures the human mind in its entirety. A first step in this direction is to create a model that can predict human behavior in a wide range of settings. Here we introduce Centaur, a computational model that can predict and simulate human behavior in any experiment expressible in natural language. We derived Centaur by finetuning a state-of-the-art language model on a novel, large-scale data set called Psych-101. Psych-101 reaches an unprecedented scale, covering trial-by-trial data from over 60,000 participants performing over 10,000,000 choices in 160 experiments. Centaur not only captures the behavior of held-out participants better than existing cognitive models, but also generalizes to new cover stories, structural task modifications, and entirely new domains. Furthermore, we find that the model's internal representations become more aligned with human neural activity after finetuning. Taken together, our results demonstrate that it is possible to discover computational models that capture human behavior across a wide range of domains. We believe that such models provide tremendous potential for guiding the development of cognitive theories and present a case study to demonstrate this.

Problem

• Creating a unified computational model of human cognition
• Predicting human behavior across diverse experimental settings
• Aligning model representations with human neural activity patterns

Innovation

• Finetuned state-of-the-art language model
• Large-scale Psych-101 dataset training
• Predicts human behavior universally