Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster D in Poster Session D: Wednesday, August 5, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Using recurrent neural networks to examine sequential effects in risky choice

Ziheng Feng¹, Carlos G. Correa¹, Marcelo G Mattar¹; ¹New York University

Presenter: Ziheng Feng

Many models of human risky choice assume each decision is independent, leaving the role of sequential feedback unexamined. We extend Peterson et al's. (2021) Neural Expected Utility framework with a Gated Recurrent Unit (GRU) to capture how feedback history shapes subjective valuation across repeated trials. The recurrent model outperforms the static baseline (BCE: 0.599 vs. 0.645), with gains persisting even at a single hidden dimension, confirming that temporal structure rather than parameter capacity drives the improvement. Behaviorally, consecutive wins induce increased sensitivity to EV. Unlike the static baseline, the GRU dynamically captures this shift, demonstrating how feedback modulates subjective valuation.

Topic Area: Decision-Making, Cognitive Control & Event Cognition