Automated Scientific Discovery of Mind and Behavior

2026-02-14 · Source: Ai2 · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Advanced, extended

Summary

Sebastian Mistake, a professor at Osnabrück and Brown, presented an update on Tora, an AI system designed for automated scientific discovery in cognitive science. The system addresses the problem of theoretical fragmentation in cognitive control research, where numerous specific effects and small models exist without integration. Tora aims to expand the space of models and experiments, accelerate scientific search, and document the entire scientific process, including thought processes. The system operates as a closed loop, with an artificial scientist iterating through experimentation on human participants and model inference. Mistake demonstrated Tora using a two-armed bandit experiment, where the system autonomously designed an experiment, collected data, and discovered a reinforcement learning model. He highlighted Tora's ability to find models that outperform human-made ones in fit and complexity, and to express individual differences in equation terms, not just parameters.

Key takeaway

For AI Researchers and Research Scientists working on automated scientific discovery in empirical fields, you should consider adopting a closed-loop AI system like Tora to address theoretical fragmentation and accelerate model discovery. Your efforts should focus on integrating diverse sub-problems of the scientific process, such as experimental design and model inference, into a unified framework, potentially leveraging knowledge graphs to represent relational scientific content and ensure robustness.

Key insights

AI can overcome human cognitive limitations to accelerate and integrate scientific discovery in complex fields like cognitive science.

Principles

• Scientific content is relational.
• Model search can be framed as program search.
• Individual differences extend to cognitive mechanism structure.

Method

Tora combines recurrent neural networks (RNNs) to infer latent dynamics from behavioral data, followed by symbolic regression (e.g., using Cindy) to derive interpretable cognitive equations from the RNN's scalar latent variables.

In practice

• Use Tora to automate cognitive science experiments.
• Employ RNNs for latent dynamic inference.
• Apply symbolic regression to derive interpretable equations.

Topics

• Automated Scientific Discovery
• Cognitive Science Research
• Tora System
• Model Discovery
• Knowledge Graphs

Best for: AI Researcher, Research Scientist, AI Scientist

Related on AIssential

• 🔬 Automating Science: World Models, Scientific Taste, Agent Loops — Andrew White — AI agents like Cosmos are automating scientific discovery by integrating literature, data analysis, and experimental feedback loops.
• A Three-Layer Framework for AI in Scientific Discovery — True scientific discovery in AI hinges on qualitative reasoning for model formation, transcending mere search or execution.
• Can an AI agent run the entire scientific method without human supervision? — Arbor introduces a hypothesis tree to enable AI agents to accumulate knowledge and conduct cumulative, strategic autonomous research.
• MIND: AI Co-Scientist for Material Research — MIND is an LLM-driven multi-agent framework for automated, experimentally verified hypothesis validation in materials science.
• Four ways Google Research scientists have been using Empirical Research Assistance — ERA accelerates scientific discovery by generating expert-level empirical software for diverse, complex real-world problems.
• Towards Generalist Agents for Accelerating Scientific Discovery171 — LLMs can accelerate scientific discovery by generating and refining hypotheses within agentic, iterative workflows.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Ai2.