Gouki Okazawa: “Modeling integration of dynamic multi-dimensional sensory evidence for perceptual decision-making”

March 12, 2018  12:15 〜 13:25

CiNet 1F Conference Room

Gouki Okazawa

Center for Neural Science, New York University, USA

Host : Hiromasa Takemura (Amano group)


Perceptual decision-making is a process of commitment to a plan of action based on sensory evidence gathered from the outside world.
Sensory evidence is multi-dimensional and dynamically changing over time; visual signals, for example, consist of spatiotemporal patterns of information and the decision-making process converts these spatiotemporal patterns to an action. In this talk, I will discuss how one can gain insight into this conversion mechanism from behavioral data using psychophysical reverse correlation. First, through computational modeling, I show that psychophysical reverse correlation reflects the complexity of both sensory and decision processes and that one needs a detailed, quantitative model to draw a valid conclusion from the reverse correlation. Second, based on this framework, I show that empirical data obtained from a face discrimination task could be explained by linear spatiotemporal integration of evidence conferred by individual facial features. Together, I propose that one can leverage psychophysical reverse correlation and quantitative behavioral modeling to understand the conversion of sensory signals to an action in perceptual decision-making.