Fig. 3: Decision-making model comparison and parameter distribution: addictive condition.

a,b, For all models, ΔBIC was defined as the difference between each model’s BIC and the best-performing BIC in the alcohol group (a) and cannabis group (b). The α-bias model performed best across groups. c,d, For the best-performing α-bias model, we performed parameter recovery by simulating data from the parameter estimate and refitting the simulated data in the alcohol group (c) and the cannabis group (d). Pearson correlation between original and recovered parameter estimates was used to perform significance testing. All parameters (α, β, φ) displayed excellent parameter recovery across groups (P < 0.01 across parameters and groups). e,f, Distributions of parameters were extracted from the α-bias model for the alcohol group (e) and cannabis group (f). The left panels display the joint distributions of α (learning rate) and φ (modulation factor) as these interact directly in the model, while β (inverse temperature) is visualized separately. A two-sided one-sample t test without correction was used for all significance testing. In the alcohol group, φ was found to be significantly positive (t = 2.159, P = 0.034), while in the cannabis group, it was found to be significantly negative (t = −5.590, P < 0.001). g, Base learning rate (α) was higher in cannabis users versus alcohol users (t = 4.90, P < 0.001). h, There was no difference between groups for inverse temperature (t = 0.521, P = 0.604). i, Modulation parameter φ was significantly higher in alcohol users versus cannabis users (t = 6.01, P < 0.001). A two-sided independent-sample t test without correction was used for all significance testing for g and h.