The game changed, what about a new strategy?

Etienne Koechlin presented today his recent work on a Computational Theory of Prefrontal Executive Control. The experiment which they target is about the strategy change: what happens in the prefrontal when a participant is instructed to tackle multiple task-sets requiring, the same, new, or past strategies? The latter one is called recurrent trials. The participants were informed about their performance by feedbacks and external cues. Different task sets vary in time and, supposedly, complexity.

His student and he have proposed a model combining reinforcement learning and baysian inference to explain how humans can either invent a new strategy or use past strategies. They used 3-order baysian inference model to capture the memory of past strategies.