From the Lab of Thomas Serre
Associate Director of the Center for Computational Brain Science, Director for the Center for Computation and Visualization, Professor of Cognitive, Linguistic and Psychological Sciences, Professor of Computer Science
Robert J. and Nancy D. Carney Institute for Brain Science
Concept Recursive Activation FacTorization for Explainability (CRAFT)
Why it’s Important
CRAFT is an important tool because it can help researchers verify AI systems – or debunk them. Trust and transparency are essential when, increasingly, AI technologies are used to make critical decisions in health care, education and finance.
What it Does
Computer vision systems are only as good as the training sets that they're built on. Biases in training data can cause errors. For example, AI systems might mistake a fisherman for a fish because the images of fish the system was trained on had fishermen holding them. The CRAFT method not only identifies where AI systems are focused while studying images, but what systems see and how they use that information to make decisions.
Where it’s Used
The scientific paper describing CRAFT has been cited more than 100 times since its 2024 publication and the tool has been used by companies and government agencies working on projects ranging from fire detection to electricity generation to aerospace design.