Seongmin A ParkI am a project scientist in the team of Erie Boorman at the center for mind and brain in the University of California, Davis. I received my Ph.D. in Culture Technology with Jeaeseung Jeong at KAIST. I was a postdoctoral researcher in the team of Jean-Claude Dreher at CNRS. In my research, I strive to investigate how the human brain implements reinforcement learning (RL) and decision making (DM) while combining with concepts and tools proposed by animal models, behavioral economics, and machine learning. By incorporating behavioral experiments, neuroimaging (fMRI and EEG), and computational models, my research addresses how the brain transforms the complex experiences into low-dimensional representation of abstract knowledge and further promotes behavioral flexibility in ever-changing environments.
Contactseongmin.a.park [at] gmail.com
267 Cousteau Pl, Davis, CA 95618, USA
|
There have been long-standing interests in the brain’s ability to construct map-like representations that encode relationships between entities, concepts, events, and even abstract task states. The exciting findings on the hippocampus (HC) and entorhinal cortex (EC) support this idea of ‘cognitive maps’. These findings demonstrate that this system organizes not only spatial but also non-spatial relational information to guide goal-directed decision making. Having a cognitive map affords the brain the ability to infer unlearned relationships as if finding a novel route in a physical space, planning goal-directed decisions with sparse and delayed feedback, and generalizing previous experiences to make novel decisions that were never made before. This behavioral flexibility is a hallmark of human intelligence but remains a major challenge in artificial intelligence (AI).
The neural mechanisms for constructing cognitive maps have been primarily examined in a physical space or in continuous task dimensions where agents can readily infer their current state from continuous sensory feedbacks (e.g. visual, auditory, vestibular). On the other hand, little is known about how the brains construct a structural representation in abstract knowledge dimensions where agents often learn discrete relationships from separate experiences in the absence of explicit sensory feedback.
Considering that many everyday decisions require comparisons between discrete decision options in which we have learned their feature values in separate experiences (e.g. who to collaborate with, which team will be victorious, or who will be a better political leader), it is important to examine how the brain represents the relationships between abstract concepts or discrete entities and how subjective representation guides different decisions and judgments across individuals.
My research focuses on how the brain learns latent states in abstract task spaces, integrates separately sampled discrete relationships between entities into a coherent representation and uses the cognitive map to guide novel decisions.
Contact