Study 1 --- The Czech and Chinese sample differed in a number of characteristics (see Figure S1 and Figure S2). ![FigS1][1] **Figure S1. Differences in individual measures between Czech and Chinese samples in Study 1.** The figure shows Spearman correlation coefficients and 95% confidence intervals. Positive correlations indicate that participants from the Czech sample were higher on a given measure. ![FigS2][2] **Figure S2. Differences in values between Czech and Chinese samples in Study 1.** The figure shows Spearman correlation coefficients and 95% confidence intervals. Positive correlations indicate that participants from the Czech sample gave the value higher importance. While 28.1% Czech participants who were given the opportunity to choose a version of the task for the third round chose the BEFORE version of the task, and 26.3% chose the AFTER version of the task, 22.1% of Chinese participants chose the BEFORE version and 35.1% chose the AFTER version. The correlation between the choice of a version and nationality was however not significant, *r*_S = -.10, 95% CI [-.22, .01], *p* = .07. The difference between the number of reported correct predictions in the third round between those who chose the AFTER version of the task actively and those who were assigned to it from the control group did not differ between the two samples, *t*(244) = 1.45, *p* = .15, *b* = 0.705, 95% CI [-0.256, 1.665]. Similarly, the difference between those who chose the AFTER version and those who were assigned to it from the experimental group did not differ between the two samples, *t*(161) = 0.14, *p* = .89, *b* = 0.081, 95% CI [-1.079, 1.241]. The difference in the baseline measure of cheating between those in the experimental group who chose the AFTER version and those who chose to be assigned randomly did not differ between the two samples, *t*(236) = -0.38, *p* = .70, *b* = -0.190, 95% CI [-1.172, 0.792]. Similarly, the difference in the baseline measure of cheating between those in the experimental group who chose the AFTER version and those who chose the BEFORE version did not differ between the two samples, *t*(175) = 1.11, *p* = .27, *b* = 0.671, 95% CI [-0.525, 1.867]. There was some indication that the difference in the baseline measure of cheating between those in the experimental group who chose to be assigned randomly and those who chose the BEFORE version differed between the two samples, *t*(219) = 1.96, *p* = .05, *b* = 0.861, 95% CI [-0.007, 1.729]. The pattern of mean numbers of reported correct predictions differed between the two samples, but the difference in the baseline measure was significant neither for the Czech sample, t(121) = 1.32, p = .19, d = 0.25, 95% CI [-0.12, 0.61], *M*_random = 5.12, *M*_before = 4.74, nor for the Chinese sample, *t*(98) = -1.43, *p* = .16, *d* = -0.30, 95% CI [-0.72, 0.12], *M*_random = 5.92, *M*_before = 6.41. The difference between the baseline cheating and the number of reported correct predictions in the third block also did not differ between the two samples, *t*(96.0) = -0.97, *p* = .34, *b* = -0.362, 95% CI [-1.097, 0.373]. Similarly, there was no such difference in those who were assigned the AFTER version after choosing to be assigned randomly, *t*(65.0) = -0.66, *p* = .51, *b* = -0.281, 95% CI [-1.114, 0.551], as well as those who were assigned the AFTER version randomly from the control group, *t*(148.0) = 0.79, *p* = .43, *b* = 0.263, 95% CI [-0.387, 0.913]. Values measured by PVQ were analyzed for the two samples separately. No value consistently predicted the choice of a version in both samples (see Figure S3 and Figure S4). ![FigS3][3] **Figure S3. The association of values measured by PVQ and cheating in the baseline measure and choice of a version of the task in the Czech sample in Study 1.** The figure shows Spearman correlation coefficients and 95% confidence intervals. ![FigS4][4] **Figure S4. The association of values measured by PVQ and cheating in the baseline measure and choice of a version of the task in the Chinese sample in Study 1.** The figure shows Spearman correlation coefficients and 95% confidence intervals. Study 2 --- **Study 2 pre-registered analysis of the relationship between the baseline measure of cheating and the choice of task versions** A linear regression with the reported number of correct predictions in the baseline measurement as the dependent variable, the number of selections of the AFTER version of the task in the third and fourth rounds (coded as -0.5, 0, and 0.5), the round with the fee condition, and their interaction as predictors, showed that participants who selected the AFTER version of the task in the third and fourth rounds a higher number of times reported a higher number of correct predictions in the baseline measurement, t(497) = 13.30, p < .001, b = 2.88, 95% CI [2.45, 3.30]. **Study 2 pre-registered analysis of the relationship between the presence of a fee, round of the task and the choice of task versions** A mixed-effect logistic regression with the choice of the version as the dependent variable, the presence of a fee, round number and their interaction as predictors, showed that participants were less likely to select the AFTER version in the presence of a fee, z = -6.49, p < .001, OR = 0.34, 95% CI [0.25, 0.48], and more likely in the fourth round of the task, z = 3.50, p < .001, OR = 1.71, 95% CI [1.27, 2.31]. The difference between the fee and no-fee condition was larger in the fourth round, z = -2.49, p = .01, ratio of OR = 0.40, 95% CI [0.19, 0.82]. When the difference between rounds was analyzed separately for the two conditions, there was no difference between the two rounds for the fee condition, z = 0.29, p = .77, OR = 1.06, 95% CI [0.74, 1.51], but the AFTER version was more likely to be selected in the fourth round than in the third round in the no-fee condition, z = 4.10, p < .001, OR = 2.13, 95% CI [1.48, 3.06]. [1]: https://mfr.osf.io/export?url=https://osf.io/ytz9e/?action=download%26mode=render%26direct%26public_file=False&initialWidth=848&childId=mfrIframe&parentTitle=OSF+%7C+FigS1.png&parentUrl=https://osf.io/ytz9e/&format=2400x2400.jpeg [2]: https://mfr.osf.io/export?url=https://osf.io/qykbn/?action=download&mode=render&direct&public_file=False&initialWidth=848&childId=mfrIframe&parentTitle=OSF%20%7C%20FigS2.png&parentUrl=https://osf.io/qykbn/&format=2400x2400.jpeg [3]: https://mfr.osf.io/export?url=https://osf.io/hfnuj/?action=download&mode=render&direct&public_file=False&initialWidth=848&childId=mfrIframe&parentTitle=OSF%20%7C%20FigS3.png&parentUrl=https://osf.io/hfnuj/&format=2400x2400.jpeg [4]: https://mfr.osf.io/export?url=https://osf.io/5g7dp/?action=download&mode=render&direct&public_file=False&initialWidth=848&childId=mfrIframe&parentTitle=OSF%20%7C%20FigS4.png&parentUrl=https://osf.io/5g7dp/&format=2400x2400.jpeg