Drift dynamics interact with a confirmation bias in visual working memory
Neuroscience ePoster by Hyunwoo Gu, Joonwon Lee, Hyang-Jung Lee, Heeseung Lee, Minjin Choe, Sungje Kim and 5 more
Part of COSYNE 2023 posters, Stanford University; Seoul National University; New York University Shanghai
Friday
Abstract
Committing to a decision tends to bias future behaviors, a phenomenon dubbed “confirmation bias”. Daily life requires us to make decisions on remembered past experiences; still, the understanding of confirmation bias in working memory remains sparse. Here, we show that the stimulus-specific drift dynamics should be incorporated into any viable account of confirmation bias in working memory. We crafted a paradigm where participants view a grating briefly, make a binary decision on it, and estimate its orientation, with varying delays between those epochs. We developed a dynamical model wherein decision and drifting memoranda mutually affect each other in determining eventual mnemonic estimates. This model accounted for the key dynamic features of behavior reports and BOLD responses in the visual cortex, including (i) the decision-time-dependent stimulus-specific and decision-consistent biases and (ii) the relation between pre- and post-decision biases. Furthermore, these dynamic features of confirmation bias naturally emerged from recurrent networks trained to the above experimental paradigm, suggesting an interplay between the winner-take-all decision-making mechanism and the drift-driven working memory dynamics. The work presented here highlights drift dynamics as a novel link mediating an interplay between decision-making and working memory.
Topics
Details
- Session
- Poster Session I