CAREER: The cognitive demands of uncertainty source inference

People constantly face surprising events and must figure out what they mean. A teacher needs to determine whether a student’s poor test reflects real difficulty with the material or just a bad day, and a person needs to judge whether a partner’s withdrawn behavior signals a lasting shift in the relationship or just a passing mood. Making these judgments accurately matters for decisions in school, work, relationships, and many other domains of life, yet the computations and neural mechanisms supporting this process, and why they sometimes fail, remain poorly understood. The proposed research aims to investigate how the human brain solves this challenge, deciding whether a surprising outcome reflects a true change in the world or simply random noise in what was observed, and to test the new hypothesis that this judgment requires mental effort that is shaped by motivation and can fail when that effort is depleted. Findings may inform the design of more reliable artificial intelligence systems, which still struggle in uncertain environments, and provide quantitative tools for understanding psychiatric conditions such as anxiety, in which people often misinterpret uncertainty. The project also proposes educational activities to train the next generation of scientists at the interface of neuroscience and artificial intelligence through new courses at different levels and openly shared materials. This challenge has a specific computational structure. In real-world environments, uncertainty arises from two distinct sources that the brain must disentangle: volatility, meaning how quickly the underlying state of the world changes, and stochasticity, meaning noise in what is observed given that state. These two sources demand opposite responses, since volatility should speed up learning while stochasticity should slow it down, yet both increase the variability of observations, making them difficult to dissociate. The proposed research reconceptualizes this dissociation as an effortful executive function rather than an automatic computation, predicting that decision-making should deteriorate under cognitive load, improve with motivation, and show systematic individual differences that are exacerbated when cognitive resources are taxed. Three interconnected aims intend to test this idea through large-scale behavioral experiments examining how cognitive load and individual differences shape uncertainty source inference, functional magnetic resonance imaging examining how distributed activity in prefrontal circuits implements this inference and how motivation modulates it, and computational modeling that jointly predicts choices and their timing from a single inference process. Knowledge gained through this work may help link the computational problem of uncertainty source inference to neural mechanisms of cognitive control, provides new tools for assessing maladaptive decision-making and uncertainty processing in psychiatric conditions, and offers principles for designing artificial intelligence systems that handle uncertainty more robustly. 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. NSF Award ID: 2543567 | Program: 01002627DB NSF RESEARCH & RELATED ACTIVIT | Principal Investigator: Payam Piray | Institution: University of Southern California, LOS ANGELES, CA | Award Amount: $741,148 View on NSF Award Search: https://www.nsf.gov/awardsearch/show-award/?AWD_ID=2543567 View on Research.gov: https://www.research.gov/awardapi-service/v1/awards/2543567.html