Figures and data

Task design, affect model posterior predictions and model comparison.
A. The task design. B. Mean affect ratings for each rating type (engaged, happy, or confident), by distancing group, compared to model predictions (light-coloured lines). C. Example comparison of predictions from the best-fitting model (light-coloured lines) to raw affect ratings from three different individuals with the median pseudo-R2 for each rating type. D. Model fit compared to the best-fitting model (time elapsed, overall with dual learning rate) in terms of their ELPD (i.e., higher ELPD [or less negative ELPD compared to the best-fitting model] is better), estimated via Bayesian approximate LOO cross-validation47. Ribbons in B-C and error bars in D denote standard errors.

Associations between affect parameters and transdiagnostic mental health dimensions, and results of PLS regression.
A-C. Estimated differences in baseline affect (

Associations between treatments and affect model parameters.
A-B. Estimated mean differences in individuals’ baseline affect (

Parameter distributions used to simulate data and test parameter recovery.
+i.e., so β ∈ [0,10]

Parameter recovery for A) the joint RL-affect model, and B) the between-rating RL-affect model.

Comparison of Q-learning parameters and effects of distancing between dual learning rate models fit to choices alone and the joint RL-affect model additionally fit to affect ratings.

Associations between higher transdiagnostic psychiatric symptom factor scores and additional affect parameters (A-C), correlation between variance in happiness rating and compulsive behaviour score (D), and the simulated effect on happiness ratings of higher

Effects of cognitive distancing (A) and antidepressant use (B) on expected value and prediction error parameters, derived from the between-rating RL-affect model.

Associations between baseline affect and affective drift, and self-reported fatigue.