One interesting result was that, at the trial level, RL was better at predicting participants' win-stay choices while active inference was better at predicting their lose-switch choices (and all switch choices). Active inference was also a bit better at predicting when they chose the best option.
