The 14th Prediction Science Seminar was held on 24 January 2023 (Tuesday), 10:30-12:00 (JST), and was presented by Prof. Naoki Honda from Hiroshima University (right top in the picture).
 In his talk, he presented his recent work on decision-making [Konaka and Honda, 2022]. Animals make decisions in many different situations. Sometimes they make irrational choices that result from a dilemma between reward and curiosity. However, conventional decision-making models, such as reinforcement learning and the free energy principle, have failed to capture these irrational behaviors. To overcome this situation, Prof Honda's research group developed the Reward-Curiosity Decision Model (ReCU). Using the ReCU model, they were able to describe a variety of behaviors, such as greedy behavior toward rewards, information-seeking behavior with high curiosity, and conservative behavior avoiding uncertainty. They also developed the inverse free energy principle (iFEP), which is a particle filter-based method for estimating curiosity levels from behavioral data. They applied the method to actual rat behavioral data and found that the rat's negative curiosity level reflected the irrational choice of sticking to one option. Overall, they proposed a powerful model and method for analyzing the decision-making mechanism.
 After his presentation, we had a fruitful discussion about the technical details and future directions of the method.

Reference:
Konaka Y and Honda N. Decoding reward–curiosity conflict in decision-making from irrational behaviors. bioRxiv 2022.04.24.489304; doi: https://doi.org/10.1101/2022.04.24.489304