Model-Based Reasoning in Humans Becomes Automatic with Training
The weighting parameter w represents a measure of model-based (w = 1) relative to model-free (w = 0) control. w was lower in the dual-task (high load) condition compared to the single-task (low load) condition in naïve (‘high load group’, day 1) but not trained (‘low load group’, day 3) subjects. Vertical lines represent SEM. * denotes p < 0.05. α = learning rate, β = inverse temperature, ε = lapse rate. (A) Mean best-fitting parameters for day 1 of training in the ‘high load group’. (B) Mean best-fitting parameters for day 3 of training in the ‘low load group’.