ATLAS: Active Theory Learning for Automated Science

2026-06-10

AI Summary

This work addresses the challenge of automatically designing efficient experiments to discover interpretable mechanistic models of behavior. It introduces ATLAS, a novel framework that uniquely integrates active learning with interpretable mechanistic modeling by iteratively generating sparse, disentangled recurrent neural network (RNN) hypotheses and optimizing structured temporal experiments that maximally discriminate among these hypotheses. Tailored for automated theory discovery in cognitive science, the approach achieves a 5–10× improvement in sample efficiency over random or expert-designed experiments across multiple evaluation metrics and successfully recovers the intrinsic decision-making mechanisms of reinforcement learning agents.

Abstract

Advancing scientific understanding through mechanistic modeling requires posing the right experimental questions to yield maximally informative data. To automate this pursuit within cognitive science, we introduce ATLAS (Active Theory Learning for Automated Science), an active learning framework for the data-driven discovery of interpretable behavioral models. ATLAS iterates between generating mechanistic hypotheses--instantiated as a diverse ensemble of sparse neural networks (Disentangled RNNs)--and designing experiments that optimally distinguish between them. We test this approach on the problem of recovering reinforcement learning agents from their behavior in bandit tasks. ATLAS designs varied sequences of qualitatively novel experiments with temporal structure tailored to underlying agent characteristics. The models trained on these experiments are evaluated against a comprehensive set of metrics for mechanistic modeling that capture behavioral, structural, and computational similarity. ATLAS achieves a 5-10x improvement in sample efficiency across all metrics compared to random experimentation, and its performance is further validated against expert-designed experiments derived from literature. These in silico results showcase ATLAS's potential to accelerate human-interpretable insights in cognitive science and other domains where scientific inquiry relies on discovering mechanistic models.

Problem

• active learning
• mechanistic modeling
• interpretable models
• automated science
• behavioral experiments

Innovation

• active learning
• mechanistic modeling
• interpretable models
• disentangled RNNs
• automated science