Seminar Artificial Intelligence - Dr. Catherine Sibert

Title: Decoding the mind: Exploring the underlying structure of cognition with neuroimaging data

Abstract:
Efforts to understand the complex system of the mind tend to approach the problem with a bottom-up strategy, interpreting low-level neural signals, or with a top-down approach, proposing a plausible model of cognitive components and comparing model performance to observed human behavior. Research using each method tends to progress in isolation from the other, but here we combine both approaches by using a constrained cognitive architecture as a framework through which to interpret high-density, distributed brain imaging data. We found that a broad spectrum of brain activity was well predicted by the recently proposed Common Model of Cognition, a consensus, minimal architecture for intelligent agents. The Common Model is a potential starting point from which to continue refining and validating the model structure, expanding the scope of the model’s predictions, using the model to compare brain activity in various clinical populations, and applying the model constraints in the development of human-like and human-usable artificial intelligence systems.