Automated Scientific Discovery of Mind and Brain Workshop#

Artificial intelligence and automation are rapidly transforming scientific practice, where they enable unprecedented scientific discoveries–from generating novel conjectures in mathematics, to identifying novel quantum states in physics, to predicting unprobed protein structures in chemistry. These technologies not only accelerate scientific discovery, but also reshape how science is conducted. Although the fields of AI for science and automated scientific discovery have primarily advanced within the natural and engineering sciences, they are now making significant inroads into psychology and neuroscience–offering approaches for automated scientific discovery in research on mind and brain.

This two-week workshop introduces participants to emerging automated scientific discovery methods for the study of mind and brain. Participants will gain exposure to methods for automating the design and execution of behavioral experiments, the discovery of scientific models, and closed-loop scientific discovery.

Workshop Content#

Basic Methods for Automated Behavioral Research#

By completing this section, participants should be able to:

  • Construct and fit models of mind and brain with PsyNeuLink

  • Automate the generation of experimental designs with SweetPea

  • Automate the generation of web-based behavioral experiments with SweetBean

  • Automate web-based data collection with Prolific

  • Automate entire behavioral research cycles, by integrating experimental design, experiment generation, and computational modeling into a closed-loop with AutoRA

AI-driven Techniques for Automated Model Discovery and Large Language Models for Scientific Discovery.#

By completing this sections, participants should be able to:

  • Gain a basic understanding of emerging approaches to cognitive model discovery and LLM-based approaches to automating scientific discovery

  • Gain hands-on experience with general equation discovery with SINDy

  • Automate cognitive equation discovery with SPICE

  • Automate cognitive program discovery with Gecko

  • Run experiments LLM-based synthetic participants with Centaur and SweetBean