skip to main content

Title: Inferring Probabilistic Reward Machines from Non-Markovian Reward Signals for Reinforcement Learning
Award ID(s):
2106339 1552497
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
International Conference on Automated Planning and Scheduling
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. The reward-hypersensitivity model posits that trait reward hypersensitivity should elicit hyper/hypo-approach motivation following exposure to recent life events that activate (goal striving and goal attainment) or deactivate (goal failure) the reward system, respectively. To test these hypotheses, we had 87 young adults with high trait reward (HRew) sensitivity or moderate trait reward (MRew) sensitivity report frequency of life events via the Life Event Interview. Brain activation was assessed during the functional MRI monetary-incentive-delay task. Greater exposure to goal-striving events was associated with higher nucleus accumbens (NAc) reward anticipation among HRew participants and lower orbitofrontal cortex (OFC) reward anticipation among MRew participants. Greater exposure to goal-failure events was associated with higher NAc and OFC reward anticipation only among HRew participants. This study demonstrated different neural reward anticipation (but not outcome) following reward-relevant events for HRew individuals compared with MRew individuals. Trait reward sensitivity and reward-relevant life events may jointly modulate reward-related brain function, which has implications for understanding psychopathology.

    more » « less
  2. null (Ed.)