Probing Changes in Neural Task Representations from Novice to Expert
A) Modified two-step task. (Left) Subjects make a 1st stage choice in starting state S, then transition to states A or B with prob.0.7 or 0.3. Then a 2nd stage choice leads to reward or not. Good performers learn to take the green path. (Right) MB or sophisticated MF control are further discriminated following infrequent introductions of un-signaled reversals in the transition probabilities, where they differ in their predicted rate of adaptation to transition reversals. B) This plot shows preliminary data on a version of the two-step task (n=26) demonstrating significantly increased MB-like behaviorwith training when comparing sess. 4 with sess. 1. C) (Right) Simulation results show that both MB and sophisticated MF strategies navigate to the rewarding outcome (green path in A) equally well. (Left) Only sophisticated MF exhibits perseverative choice biases after a rare transition to state A (purple path).
“Practice makes perfect” it is said, but we know little about how the brain facilitates the transition from novice to expert during complex tasks. The goal of this proposal is to gain insight into how task-related computations within and across different brain areas and networks are modified as a function of increasing competence on a task. We propose that task expertise is associated with increased efficiency in how states and actions get represented in the brain: that the brain compresses both the state and action-space so as to maximize efficient and accurate task execution. To test this, we will combine computational modeling with fMRI and behavior while participants perform several tasks designed to enable us to investigate these different forms of task-relevant data compression.
