Recording
Speaker
Weiji Ma
Bio
Wei Ji Ma is Professor of Neural Science and Psychology at New York University. He received his Ph.D. in Theoretical Physics from the University of Groningen, the Netherlands, in 2001. He switched to computational neuroscience and computational cognitive science, doing postdocs at Caltech and the University of Rochester. He was Assistant Professor of Neuroscience at Baylor College of Medicine from 2008 to 2013 before he joined New York University. His lab uses behavioral experiments, computational models, and (through collaboration) neural measures to investigate how people make perceptual and cognitive decisions. Ma has worked on topics in visual decision-making, neural coding and computation, multisensory perception, working memory, metacognition, complex planning, procrastination, and mindsets. He is the lead author of the textbook Bayesian models of Perception and Action, published by MIT Press in 2023.
Ma is the Program Director of the NIH-funded Training Program in Computational Neuroscience at NYU. He founded the “Growing up in Science” mentorship series, in which scientists tell their life stories with an emphasis on doubts, struggles, and failures. He is also a founding member of the Scientist Action and Advocacy Network, which provides pro-bono science to social and environmental non-profit organizations. He co-founded the Rural China Education Foundation, which supports community-based elementary education in rural China. In 2021, Ma received the Elman Prize for Scientific Achievement and Community Building from the Cognitive Science Society, and in 2023, he received the Impact Goals Award from the Center for Advancing Research Impact in Society.
Abstract
As DeepMind has revolutionized the AI of planning in combinatorially large problems, our lack of understanding of how humans plan in such situations has come into stark focus. The cognitive science of chess, once promising, is now virtually extinct. Planning tasks that are nowadays widely used in the field don’t require much thinking ahead. I will show that it is possible to study human complex planning in tasks of intermediate complexity while maintaining experimental tractability and computational modelability. I will describe experiments on a game that we call four-in-a-row – a variant of tic-tac-toe or Go Moku. Inspired by best-first search, we built a heuristic computational model of human play in this game and fitted it to move-level data. The model predicts moves in unseen positions, decisions in unseen tasks, eye fixation patterns, mouse movements, and response times. Moreover, the model allows us to computationally characterize the effects of expertise and time pressure. Linking back to the chess literature, I will discuss how experts differ from novices in remembering game positions and move sequences. I will describe parallel results from a very large online data set, connections to development, and ongoing work on the neural basis of complex planning. Finally, I will comment on some broader themes: resisting reductionism, the use of games to study cognition, and comparisons to other species.