Fig. 4
(A) Test–retest reliability of model parameters and task measures. The red, orange, light green and dark green background indicates bad, acceptable, good and very good reliability, respectively. Circles indicate the reliabilities of model-based parameters. Triangles indicate the reliabilities of model-free task measures. Leftwards facing square brackets represent the recoverability of the parameters. (B) Convergent validity of the trial-by-trial switch probability. Correlations between participants’ average proportion of trials where they selected a different option than on the previous trial across tasks. (C) Convergent validity of the recall accuracy in the working memory tasks. (D) Convergent validity of the model parameters in the bandit tasks. Correlations between subject-level model parameters across tasks. (E) External validity of task measures and model parameters. Correlations of subject-level task measures as well as model parameters with all questionnaire scores and performance in all three working memory tasks. Note: Exploration refers to the CEI, Openness to the BIG5 Openness subscale, Cognitive and Somatic Anxiety to the respective STICSA subscales, Depression to the PHQ-9, and Negative and Positive mood to the respective PANAS subscales.
