CAREER: Neural mechanisms underlying optimal performance

University Of Oregon Eugene, Eugene OR

Investigators

Abstract

In cognitively demanding tasks, such as writing an essay or solving a puzzle, task performance fluctuates depending on one's level of stress or arousal. Performance is poor in low arousal (tired) or high arousal (agitated) states, and reaches an optimum at an intermediate arousal level, colloquially described as being “in the zone”. Although this phenomenon has been extensively investigated in both humans and other species, it is still unknown how the brain achieves its peak performance. The goal of this project is to identify the computational principles underlying how optimal performance states are achieved and maintained by the brain. Combining insights from models of behavior and neural data in animals with artificial neural networks, this project seeks to explain how cortical circuits can regulate their own dynamical properties to optimize information processing. The Yerkes-Dodson inverted-U law of psychophysics describes the relationship between cognitive task performance and an animal's state of arousal, with best performance occurring at intermediate arousal levels. This project seeks to understand the neural mechanisms that enable flexibility and optimality in cognitive performance and whether these mechanisms can be harnessed by AI systems, working from the hypothesis that the intrinsic variability produced by neural circuits is harnessed and modulated to flexibly adapt the way they process information and generate behavior. The project will proceed along three main directions. First, it will examine the behavioral signatures of optimal and suboptimal performance states during sensory discrimination as well as naturalistic foraging, and their relationship to an animal’s arousal level and movements. Second, it will elucidate how optimal performance states arise from the collective activity of populations of cortical neurons. Third, the insights obtained from biological circuits will inform the design of brain-inspired artificial neural networks capable of learning to achieve multi-tasking in a robust, fast, and efficient way. Research, education, and outreach goals will be integrated through a novel scientific communication program conveying concepts from neuroscience and artificial intelligence through web-based comics. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.