91ֱ��

Biography

I study how rewards affect human and animal decision-making, and how we keep track of time. The data my students and I collect inform how my lab models decision-making and timing circuits in the brain. We use mathematical and computational models that are as simple as possible, while still achieving enough functionality to account for critical features of behavioral and physiological data.

Our models incorporate a layer of neural control mechanisms for optimizing the performance of the underlying decision-making circuits: that is, they help these circuits maximize rewards during simulated task performance. They generate precise, quantitative hypotheses about choices, response times and brain activity that we test with experiments in human behavior and electroencephalography (EEG).

We also focus theoretically on composing these basic circuits into larger models capable of more complex behavior. Our work in this area provides a possible theoretical link between different levels of description in psychology and neuroscience. At the neural level of description, neurons and neural populations are the objects of study; we use dynamical systems, stochastic processes and "deep learning" neural networks to model them. At the psychological level of description, entities such as percepts, goals and actions are the objects of behavioral investigation; researchers often use symbolic, computational cognitive architectures known as "production systems” to model these. But how do such radically different types of description relate to each other?

To try to answer that question, we model the emergence of the psychological description-level from the neural description-level, by building subsymbolic neural networks that emulate symbolic production systems.