Sensory neuroscientists tend to think of humans and other species as making optimal decisions — those that are limited only by noise in sensory encoding. Psychologists and economists, by contrast, have spent many decades describing pernicious cognitive biases that corrupt human decisions. Which of them is right? Dr. Summerfield will argue that one simple process model, based on the sequential sampling framework but incorporating a selective, limited-capacity filter (or "bottleneck"), can explain a number of violations of axiomatic rationality — including choices that are intransitive, that fail to show independence from irrelevant alternatives, or that are susceptible to framing effects. Critically, however, the simulations show that the selective bottleneck paradoxically *increases*  decision accuracy by protecting choices against noise arising 'late', during information integration. In other words, cognitive biases may paradoxically play a key role in optimising decisions in the face of noise arising in neural computations.

Faculty Host: Michael Woodford, PhD, John Bates Clark Professor of Political Economy at Columbia University.

Registration is strongly encouraged. Seating is limited.

For questions about the lecture, please contact [email protected].

Those who wish to meet the speaker during the visit should contact Michael Woodford.

This seminar is part of the Systems, Cognitive and Computational Neuroscience Seminar Series at Columbia's Zuckerman Institute, which focuses on cognition and decision making research. Internationally renowned speakers present their recent work on these topics using behavioral, neurobiological and computational approaches. Seminars take place approximately every other week on Tuesdays at 4 pm in the Jerome L. Greene Science Center (9th floor).