Scripts on modeling Active Inference POMDPs
The goal of my thesis is to apply the POMDP framework of Active Inference to experimental explore-exploit tasks from developmental psychology. The specific tasks were conducted by Sumner et al. (2019a, 2019b) and indicate that children tend to focus on exploration rather than reward maximizing. I'm designing Active Inference agents using the pymdp package by Heins et al. (2022). The main research question is which set of model parameters to choose in order to make the agent behave like a child or like an adult.
Heins, C., Millidge, B., Demekas, D., Klein, B., Friston, K., Couzin, I., & Tschantz, A. (2022). pymdp: A Python library for active inference in discrete state spaces. arXiv preprint arXiv:2201.03904.
Sumner, E. S., Steyvers, M., & Sarnecka, B. W. (2019). It's not the treasure, it's the hunt: Children are more explorative on an explore/exploit task than adults. In CogSci (pp. 2891-2897).
Sumner, E., Li, A. X., Perfors, A., Hayes, B., Navarro, D., & Sarnecka, B. W. (2019). The Exploration Advantage: Children’s instinct to explore allows them to find information that adults miss.