Data visualization

For Experiment 1A, we divided the levels of outcome uncertainty in six percentile bins per gain/loss condition, such that the 1^st bin represents 1/6^th of the lowest levels of outcome uncertainty, the 2^nd bin represents the 1/6^th– 2/6^th of the lowest levels, etc. Similarly, absolute expected values were divided in six percentile bins (the 1^st bin represents 1/6^th of the lowest values, the 2nd bin the 1/6^th– 2/6^th of the lowest values, etc.). We did the same for Experiment 1B, 1C and for Experiment 2, except that the outcome uncertainty levels and the absolute expected values were divided in eight percentile bins per gain/loss condition. For each bin, we calculated a mean curiosity rating or mean percentage willingness to wait for gains and losses separately. Next, these mean scores per bin were averaged across participants and the standard error of the mean (SEM) was calculated for each bin of outcome uncertainty and absolute expected value of the gain and loss trials separately. Least squares lines illustrate the effects (Fig 2).

In all panels, the x-axis depicts percentile bins of the levels of outcome uncertainty (left) and the absolute expected values (right) for both gains (in red) and losses (in blue). The y-axis depicts the mean curiosity ratings for each percentile of outcome uncertainty and absolute expected value in the gain and loss context (A, B and C) or the % of trials in which participants were willing to wait for the outcome in the gain and loss context (D). The error bars depict the standard error of the mean (SEM) for each outcome uncertainty and absolute expected value bin of the gains or loss trials. For other details of the data visualization, see Methods – 2.5 Data Visualization. A. Experiment 1A showed that curiosity was higher for gains than for losses. Curiosity monotonically increased with increasing outcome uncertainty, which was not modulated by gain/loss outcome valence. There was no significant modulation of curiosity by (absolute) expected value for gains and losses. B. Experiment 1B showed that curiosity was higher for gains than for losses. Curiosity monotonically increased with increasing outcome uncertainty, which was not strongly modulated by gain/loss outcome valence. There was a significant modulation of curiosity by expected value for gain trials, but not for loss trials. C. Experiment 1C showed that curiosity was higher for gains than for losses. There was a monotonic increase of curiosity with increasing outcome uncertainty, which was not modulated by gain/loss outcome valence. There was a significant modulation of curiosity by absolute expected value for gain and for loss trials, such that participants were more curious about higher gains and losses. D. Experiment 2 showed that willingness to wait was higher for gains than for losses. Willingness to wait increased monotonically with outcome uncertainty, and there was no evidence for a modulation of this effect by gain/loss outcome valence. People were more willing to wait for higher compared with lower gains, but less willing to wait for higher compared with lower losses.

To illustrate to what extent individual participants showed sensitivity to gain versus loss trials, we calculated the mean curiosity ratings (Experiment 1A, 1B and 1C) and mean percentages willingness to wait (Experiment 2) for gain and loss trials separately per participant (Fig 3—left panels). Additionally, to illustrate to what extent participants were sensitive to the effect of outcome uncertainty, we calculated the difference in curiosity/willingness to wait scores between high and low outcome uncertainty for gain and loss trials separately per participant (Fig 3—middle panels). The same was done to illustrate the sensitivity to the absolute expected value effects by calculating the difference in curiosity/willingness to wait scores between high and low absolute expected value for gain and loss trials separately per participant (Fig 3—right panels).

The left panels depict individual variability in participants’ curiosity about gains compared with losses. In all panels, the x-axis depicts the mean curiosity/willingness to wait for the loss trials and the y-axis depicts the mean curiosity/willingness to wait for gain trails. Every dot depicts one participant. In all panels, most dots are above the diagonal, indicating that a great majority of the participants were more curious (A, B and C—left panels) and more willing to wait (D—left panel) for gains compared with losses. The middle panels depict individual variability in the extent to which participants were driven by outcome uncertainty for gain and loss trials. In all panels, the x-axis depicts the mean curiosity/willingness to wait for high minus low outcome uncertainty in loss trials, with positive and negative values indicating a positive or negative relationship between outcome uncertainty and curiosity / willingness to wait in loss trials respectively. The y-axis depicts the mean curiosity / willingness to wait for high minus low outcome uncertainty in gain trials, with positive and negative values indicating a positive or negative relationship between outcome uncertainty and curiosity / willingness to wait in gain trials respectively. Every dot depicts one participant. In all panels, the majority of participants have a positive relationship between outcome uncertainty and curiosity/willingness to wait for both gain and loss trials. The right panels depict individual variability in the extent to which participants were driven by absolute expected value for gain and loss trials. In all panels, the x-axis depicts the mean curiosity/willingness to wait for high minus low absolute expected value in loss trials, with positive and negative values indicating a positive or negative relationship between absolute expected value and curiosity / willingness to wait in loss trials respectively. The y-axis depicts the mean curiosity/willingness to wait for high minus low absolute expected value in gain trials, with positive and negative values indicating a positive or negative relationship between absolute expected value and curiosity/willingness to wait in gain trials respectively. Every dot depicts one participant. For Experiment 1A (A—right panel), most participants are clustered around 0, explaining the lack of an effect of absolute expected value in both gain and loss trials. In Experiment 1B (B—right panel), most participants have a positive relationship between absolute expected value and curiosity in the gain trials (as indicated by the positive y-values), whereas the relationship between absolute expected value and curiosity in the loss trials is more variable (as indicated by both negative and positive x-values). This explains the finding that participants were more curious about higher values of absolute expected value in gain trials and the lack of an absolute expected value effect in loss trials. In Experiment 1C (C–right panel), however, most participants have a positive relationship between absolute expected value and curiosity in gain trials (as indicated by positive y-values) as well as in loss trials (as indicated by positive x-values). This explains the finding that participants were more curious about higher absolute expected values in gain and loss trials. For Experiment 2 (D—right panel), most participants have a positive relationship between absolute expected value and willingness to wait in the gain trials (as indicated by the positive y-values). However, a majority of these participants have a negative relationship between absolute expected value and willingness to wait in the loss trials (as indicated by the negative x-values). This explains the finding that participants are more willing to wait for higher gains, but less willing to wait for higher losses.