Analysis Script: UNIANOVA responsible1 BY good1 /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /EMMEANS=TABLES(OVERALL) /EMMEANS=TABLES(good1) /PRINT=DESCRIPTIVE /CRITERIA=ALPHA(.05) /DESIGN=good1. **Results** Differential Dropout: To ensure that the initial condition did not adversely affect dropout, we ran a probit regression with finishing as the DV and whether the first vignette was for good or bad behavior. We found no evidence of differential dropout (completion rate .498, initial condition effect b = .028, p > .654). Therefore, our results are unlikely to be driven by differential dropout. First 750: As predicted, we found an effect of moral valence on whether people should be held responsible for actions caused by outside forces (a brain tumor). Those who perfumed good acts because of a brain tumor were held more morally responsible for those actions (M = 2.8811, SD = 1.303) than if they had performed bad acts (M = 2.413, SD = 1.229, F (1, 748) = 25.609, p < .001, d = .37, 95%CI = .225 to .514). As we failed to find a relationship between the meek subscale of Extraversion and moral responsibility (p > .712), we did not include it in our analysis. Second 750: We replicated the effect found in the first 750 collected data. Those who performed good acts caused by a tumor were still held more morally responsible (M = 2.738, SD = 1.3) than if they had performed bad acts (M = 2.407, SD = 1.248, F(1, 748) = 12.648, p < .001, d = .26, 95%CI = .116 to .404). Though numerically larger, the magnitude of the effect was not statistically larger in the first vs. the second 750 (F(1, 1496) = 1.089, p > .296). We again failed to find a relationship between Extraversion:Meek and ratings of moral responsibility (p > .406). Full 1500: Mirroring the results from the first and second replications, pooling the data showed a large effect of asymmetry in judgments of moral responsibility when the agent is not the cause of the behavior. If a brain tumor causes someone to perform morally good acts, they are still held responsible for those behaviors (M = 2.808, SD = .751) but are held less responsible if they had been caused to perform bad acts (M = 2.41, SD = 1.237, F(1, 1498) = 36.877, p < .001, d = .314, 95%CI = .212 to .415). Contrary to the previous literature and our own pilot testing, the meek subscale of the Extraversion scale showed no relationship to ratings of moral responsibility (p > .422). Secondary Analyses Could this be driven by differences in identity? We tested whether participants believed the brain tumor changed who the actor really was. In the first vignette, participants were most likely to believe the “real” actor was both the one acting with the brain tumor and without (64% or responses). This was more likely to be chosen when the tumor was making people carry out good acts (predicted probability = 67.9%, 95%CI = 64.6 to 71.2%) than bad acts (59.8%, 95%CI = 56.3 to 63.3%). Participants were next most likely to believe the “real” actor was the one acting without the brain tumor (29%). This was more likely, however, when the tumor was making the actor perform good (36.2%, 95%CI = 32.7 to 39.6%) rather than bad acts (21.8%, 95%CI = 18.9 to 24.8%). Finally, the least chosen option was that the “real” actor was the one acting with the brain tumor. This was more common, however, when the tumor was making the actor perform good (10.3%, 95%CI = 8.1 to 12.4%) rather than bad acts (4%, 95%CI = 2.6 to 5.4%; all ps for all contrasts > .05). Overall, performing bad acts caused by a brain tumor changed the identity of the actor. The individual acting immorally was less likely to be their “real” self if they had a brain tumor than if they had been performing praiseworthy acts. Bringing out the Best: In the vignettes, the behavior of the individual was completely caused by the brain tumor. It is possible that people believed that the brain tumor ‘brought out’ the good or bad that was always dormant in the individuals. To test this, we asked individuals to what extent they believed the tumor could have brought out the good or bad in the actor. Again, we found a good/bad asymmetry, but this time in the unexpected direction. People believed the tumor brought out the bad to a larger extent (M = 5.036, SD = 1.467) than it brought out the good (M = 4.434, SD = 1.526, F(1, 1498) = 60.644, p < .001, d = .402, 95%CI = .3 to .504). Second Vignette All participants read both the good and bad vignettes in random order. When it came to judgments in the second vignette, we found no differences between measures of moral responsibility between good and bad acts (F(1, 1498) = 2.611, p > .106). Instead, all participants found the actor not very responsible for their behavior. In addition, there were no differences in average responsibility ratings between the first and second vignettes (F < 1). We also found no effect of behavior on who participants thought the “real” actor was (χ²(2) = 4.791, p > .09), in the second vignette. In general, participants believed the actor was the “real” person both before and after they had the tumor (see Table 1). Table 1: Judgments of who the "real" actor was from vignette 2. No differences in response categories between good or bad actions Response N The “real” actor was acting with the tumor 94 The “real” actor was acting both with and without the tumor 960 The “real” actor was only acting after the tumor was removed 446 Furthermore, we did not find any evidence that people thought the tumor “brought out” the good or bad in the actor. Results were the same for whether the tumor was causing bad behavior or good behavior (F < 1). Thus, all of the findings in the first vignette were not replicated within-subjects in the second vignette. The fact that the moral asymmetries found in the first vignette disappear in the second can be interpreted as a ‘correction’ effect. For both the responsibility and “brought out’ variables, the second vignette not only showed no moral asymmetry, but also was numerically at the mean of the first vignette. Thus, participants decreased judgments if they were high in the first vignette, and increased judgments if they were low. Exploratory Analyses: In our exploratory analyses, we wished to further investigate why participants may be showing a moral asymmetry in the first vignette but not in the second one. To do this, we conducted three sets of analyses. In the first, we tested whether the effects of moral asymmetry may occur by testing whether cynicism moderated the relationship between moral responsibility and valence of behavior. In the second analysis, we explored whether the asymmetry is not seen in the second vignette because seeing the ‘good’ scenario first activates certain mental processes. This was done by testing whether a) participants answered the ‘bad’ scenario quicker than the ‘good’ in the first vignette; and b) directly testing whether answering the ‘good’ scenario activates process II thinking with a test from the cognitive reflection test (CRT). Is this a function of believing people are good at their core? People, on average, believe others are good ‘by their nature’ (Newman et al., 2014). Therefore, the first moral asymmetry could occur because people think the good actor would have behaved in a morally good manner anyway (regardless of the brain tumor) and therefore should be given credit. Bad actions are not ‘normal’ and thus can be blamed on the brain tumor, exonerating people from responsibility. We tested this first by confirming the cynicism scale we used conformed to a 1-factor scale. Using Confirmatory Factor Analysis with Maximum Likelihood estimation, we did confirm a 1-scale solution that fit the data well (CFI = .992, SRMR = .017, RMSEA = .04). All loadings were significant and in the right direction (see table 1). Table 2: Factor loadings of the cynicism scale Cynicism Scale Item Factor Loading Most people would tell a lie if they could gain by it. .785 When the chips are down, most people would behave dishonestly. .72 Most people would cheat on their income tax if they had a chance. .72 If most people could get into a movie without paying and be sure that they would not be seen, they would. .709 The average person is conceited. .545 Most people are honest because they are afraid of getting caught. .34 Therefore, we extracted a single factor score using the Bartlett method and used that in subsequent analyses for testing whether the results were moderated by cynicism. We found no evidence, however, that the results were moderated by cynicism (b = -.076, p > .207). Therefore, we can conclude the results were equally robust for people who do not believe people are naturally good. Does seeing the good vignette prime process II thinking? We investigated this possibility in two ways. The first was to see if participants answered the ‘bad’ vignettes faster than the good vignettes in the first instance; this time of processing has been used before to demarcate processing styles (e.g. Rand et al., 2012). The second method was to directly test process II thinking by means of the most indicative item in the Cognitive Reflection Test. We reasoned that if participants were more likely to get this item correct after seeing the good vignette first, that would stand as evidence that the vignette was priming process II thought. We found no evidence for either hypothesis. For completion time, participants took equally as long to move on to the next page if they saw the good scenario (M = 103.904 seconds) than if they saw the bad scenario (M = 87.454 seconds, p > .538). For the CRT, participants were equally likely to get the answer correct. The predicted probability of getting the ‘widgets’ item correct was the same for participants who saw the good scenario first (21.4%) than if they saw the bad scenario first (20.1%, p > .649). In addition, the probability of getting the widgets item correct was not different seeing it between the vignettes or after, nor was there an interaction between placement and valence of vignette (both ps > .142). Therefore, we are confident the vignettes are not priming process II thinking. We found that people believed a brain tumor causing aberrant behavior was more likely to bring out underlying aspects of people if the actions were good than if they were bad. Thus, the asymmetry may be due to people believing the actor would have behaved as such even if they didn’t have a neurological disorder causing them to act that way. Since the counterfactual would have resulted in the same positive behavior, people may reason that the actor deserves credit nonetheless. Mediation by Identity: To test the possibility that this change in who the “real” actor was mediating the asymmetry, we conducted a mediation analysis. We first dummy-coded the “real” category and ran a binary mediation using percentile-corrected bootstrapped confidence intervals. The results of this model confirmed what we have seen. The overall effect of the model (b = .18, 95%CI = .127 to .234) shows that part of the reason for the asymmetry involves people believing the ‘real’ actor was good. Thus, we can see that if an individual acts in a manner that is not under their control, people largely believe it does not change their identity. However, the extent that this is true is dependent on whether the circumstances (in this case, a brain tumor) makes them act in moral or immoral ways. If people act immorally, they are more likely to be seen as someone different, their actions not attributable to who they “really” are, and are thus less morally responsible for their acts. If they are caused to act morally, people may reason that the individual would have acted that way anyway, they are thus deserving of credit.