Neural-COGNITIVE MODELING

Detailed neural models have been produced of some brain regions, most notably visual pathways, prefrontal cortex, basal ganglia, and hippocampus. Functional AI models have more broadly modeled human cognition at a functional or behavioral level. But large scale neural modeling of the dynamic interplay between many brain regions, as a human makes decisions in a challenging situation, is an elusive goal that CNES has begun to achieve.

One such program, ICArUS, is described in the sidebar and illustrated in the figure.

The ICArUS-MINDS program is producing a detailed neural model of the integrated activity of 8 brain regions while humans try to make sense of uncertain, possibly contradictory data.

Real-world problems elicit activity in many parts of the brain, some of which are not yet well understood. Competing theories are based on rodent and primate data so some of the mechanisms of human conscious decision-making are still being debated. We’re studying these processes in the context of the ICArUS program:

  • Top-down, bottom-up processing
  • Neuromodulation during cognition
  • Genesis of cognitive biases and heuristics
  • Memory effects and consolidation
  • Spatial and numerical processing
  • Semantic and hierarchical learning
CNES studies the literature and constructs detailed neural simulations including the most likely theories, and tests the model’s behavior that emerges against reported human behavior. Since we are working at a scale that few have attempted, in some cases we must come up with possible mechanisms for gaps in the research, resulting in predictions that can be tested in the laboratory and add to current understanding of the brain.  In a very real sense, we are pushing the boundaries of the science of human cognition.

Sections

Program and Sponsor

Project: Integrated Cognitive-Neuroscience Architectures for Understanding Sensemaking (ICArUS)
Sponsor: Intelligence Advanced Research Projects Activity (IARPA) Office of Incisive Analysis
Partners: Colorado University Boulder, Carnegie Mellon University, University of California Irvine, Palo Alto Research Center, University of Texas Houston, eCortex