Figure 4
From: Judgments of effort exerted by others are influenced by received rewards
Performance-related and reward-related effort estimates for self-and other-judgments, used in computational models. (A–C) Scatter plots showing the data of one representative subject. See Supplementary Fig. S4 for the data of all participants. (A) Relation between the number of key presses and effort ratings for self- (left) and other-judgments (right), based on the trials in which effort was rated before the reward was revealed (30% of all trials). The regression lines show βk for this participant. (B) Similar to (A) for the relation between trial duration and effort ratings. (C) Similar to (A) for the relation between reward magnitude and effort ratings. Note that this relation effectively captures the correlation between reward and effort ratings built-in by the task: more difficult trials were associated with higher rewards and were subjectively rated with higher efforts, therefore reward was related to the subjective estimations of exerted effort in these trials. (D) Group average ± standard error of the mean of the relationship between effort ratings and key presses (βk), elapsed trial duration (βt) and reward (βr), for self- and other-judgments. Note that the relationships between all task variables (key presses, time and reward) and effort ratings are higher for self-judgments, which is expected as in general people have more reliable estimates of their own effort. (E) Group averages ± standard error of mean of \({{\rm{\sigma }}}_{p}^{2\,}\), \({{\rm{\sigma }}}_{r}^{2\,}\) (derived from Eqs. 2–5) and ωr (derived from Model 4). In line with panel D, \({{\rm{\sigma }}}_{p}^{2\,}\) and \({{\rm{\sigma }}}_{r}^{2\,}\) values were higher for other- compared to self-judgments. This indicates that all the task variables were less reliable predictors of effort judgments for others compared to self-judgments. However, this effect was more pronounced for the performance-related criteria (key presses and time) than for rewards (two-way repeated measures ANOVA with factors self/other and performance/reward): the main effect of self vs. other F(1,50) = 16.743, p = 0.0002; the interaction F(1,50) = 16.743, p = 0.0002). Therefore, reward was a relatively more reliable predictor of effort-ratings for other- than self-judgments. Thus, the model-based reward weighting parameter (ωr) predicted by the Bayesian average model was higher for other- than self-judgments, in line with model-free results shown in Fig. 2.
