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
From Clicks to Conflict: A Computational Study of Approach-Avoidance Conflict Task Engagement
Franziska Usée1, Christiane A. Melzig2, Dirk Ostwald3, Anne Collins4; 1Phillips-Universität Marburg, 2Philipps-Universität Marburg, 3Otto von Guericke University Magdeburg, 4University of California, Berkeley
Presenter: Franziska Usée
Approach-avoidance conflicts (AAC) arise when the same action carries both appetitive and aversive consequences. Such conflicts are central to anxiety-related maladaptive avoidance; yet, the computational mechanisms underlying avoidance behavior remain poorly understood. To contribute to this research gap, we adapted a foraging-based AAC online task: participants searched visual search fields using mouse clicks for fictive coins (i.e., rewards) while risking aversive screams (i.e., punishments). Each field was associated with an explicitly instructed combination of reward and punishment magnitudes (low/high reward × low/high punishment), and participants decided whether (i.e., commitment to search) and how long to search (i.e., extent of search). Across 46 participants, the observed engagement was maximal under high reward/low punishment and minimal under low reward/high punishment. Overall, both the commitment to search and the extent of search declined over trials. Agent-based behavioral modeling showed that commitment to search was best captured by additive reward and punishment sensitivity, baseline approach bias, and trial-dependent drift, whereas the extent of search additionally benefited from a reward-punishment interaction term. Exploratory individual-difference analyses linked approach-bias estimates to self-reported approach motivation and reward valence. These findings suggest that instructed AAC behavior reflects separable contributions of valuation, motivational bias, and temporal drift.
Topic Area: Decision-Making, Cognitive Control & Event Cognition