Comment by mrvmochi

I wonder what would happen if we used RL to minimize the user's cognitive debt. Could this lead to the creation of an effective tutor model?