ePoster, FENS Forum 2024
Adaptation of rats and humans to a volatile hidden Markov model for reward collection
- Messe Wien Exhibition & Congress Center, Vienna, Austria
- Cognition, Behavioral Neuroscience and
Abstract
Animal behavior results from the integration of sensory stimuli and internal models of the functioning of the environment. Adaptation hinges on optimizing actions for survival: maximizing rewards, and minimizing punishments. For this project, humans and rats were tested in non-sensory tasks, where informed gambling strategies enhance reward outcomes.Our two-alternative forced-choice foraging task involves rats initiating trials with a central nose-poke, followed by withdrawal and selection of a water reward spout based on the estimated reward likelihood on either side. Manipulating the probability (p) of the baited side matching the previous trial (Markov model, two states: right/left, see Image), we observed rats adapting their left/right transition likelihood depending on p. In conditions with p=0.8 and p=0.2, rats developed distinct strategies: "win-stick lose-switch" for p=0.8 and "win-switch lose-stick" for p=0.2, optimizing reward collection. While post-win behavior was consistently optimal, post-lose behavior was less consistent, hinting at the challenge of applying counterfactual reasoning. Human subjects, performing a parallel task with monetary rewards, surprisingly exhibited similar adaptive strategies, even with explicit trial-by-trial feedback on counterfactual outcomes. Flexibility in adapting to changing probabilities within a session was evident in both species, with unexpected outcomes gaining greater significance.To explore decision-making further, we plan to combine this foraging task with a concurrent perceptual tactile task, unraveling the integration of congruent or incongruent forces. Neuronal recordings in rats' prefrontal areas during task execution will provide insights into the underlying mechanisms.