A Social Values Analysis of Self–Other Differences in Decision Making Involving Risk

Abstract Three experiments tested our social values analysis of self–other differences in decision making under risk. In Experiment 1, we showed that people make riskier decisions for others in domains where risk taking is valued but not in those where risk is not valued. Experiment 2 documented that it is considered more inappropriate to make a risk-averse decision for another person than for oneself in situations where risk is valued. Experiment 3 showed that self–other differences in decision making occur even when there are no self–other differences in prediction and for decisions made for a typical student as well as for a friend. We use these results to argue that decision making for others is based predominantly on the perceived value placed on risk, leading to a norm for how to decide for others in situations where such a social value exists. ACKNOWLEDGMENTS We thank Wändi Bruine de Bruin, YoonSun Choi, and Catherine E. Seta for helpful discussions regarding the ideas incorporated in this article and Andrew M. Parker for his help in coordinating the data collection process at Virginia Tech. Notes 1There is some evidence that certain conditions, such as reciprocal decision making (see Teger Zaleska for example, Slovic, Weinstein, and Lichtenstein (Citation1967) found no self–other differences in reciprocal decision making, regardless of whether the decisions were made for members of the same or opposite sex. Thus, the majority of research on monetary decisions, and to our knowledge all of the work that has used the same conditions as those of Stone et al. (Citation2002; nonreciprocal decision making for same-sex others), has found no self–other differences (Cvetkovich, Citation1972; Teger n = 56 for the ideal self condition. 3In one of the high-impact scenarios of one version, one of the names was unintentionally given as doing the risk-taking and risk-averse behavior. The results for the high-impact scenarios were reanalyzed without this item, and the results were very similar. Thus, all reported results include this item. Note. Standard deviations are in parentheses. Sample size is 52 in each condition. Note. Standard deviations are in parentheses. Sample size ranges from 25 to 27. 4This expectation regarding how to decide for others, according to our account, reflects a norm regarding how to decide for others resulting from the value placed (or not placed) on risk. Although the focus of this work is on the value placed on risk, we do not mean to suggest that is the only reason for how these norms could come about. Indeed, in at least some of our scenarios, these norms could have come about for reasons other than the value placed (or not placed) on risk, such as "It is not appropriate to tell a female to buy a Valentine Day's gift for a male she is not dating because men are supposed to initiate relationships" (Scenario 4 from Experiment 2) or "It is not appropriate to tell a friend to stay with your group of friends rather than to meet someone new because it's self-serving" (Scenario 1 from Experiment 2). Given the large correspondence we have found between situations that produce self–other differences and those where risk is valued, we believe that the value placed on risk is responsible for the development of the norms in most situations. Nonetheless, other factors are clearly at work as well, and we discuss some of these factors in the General Discussion section Note. All predictions are relative, for example, low denotes less risk taking than in the other conditions, not low in any absolute sense. For both the misprediction and self-expansion accounts, the table includes the prediction for the condition that was added to explicitly test that theoretical account against the social values account. 5As discussed by Siegrist et al. (Citation2002), Hsee and Weber (Citation1997) never discussed the reasons for why their findings were different from the earlier ones on choice dilemmas. We suggest that the difference in results occurred because of differences in the domain studied. If one assumes that risk is valued in most of the choice dilemmas scenarios but not in those studied by Hsee and Weber (as our Experiment 1 showed that risk is not valued in the monetary domain), then the different pattern of results follows under the assumption that people expect themselves to act in a more socially valued manner than they expect others to act. 6Prior to choosing the nine scenarios with the highest and lowest life impact ratings, we also eliminated one scenario that we deemed excessively long for the mode of administration (on the computer) that we were using. 7E-prime is made by Psychology Software Tools, Inc., Pittsburgh, PA; www.pstnet.com. 8In addition, participants completed some additional questions for more exploratory purposes. Specifically, they responded to the six money scenarios used in Experiment 1 as well as to some individual difference measures. The results with the money scenarios provided one significant difference, whereby people predicted that others would make more risk-taking decisions than they made when deciding for others. This result somewhat bolsters our claim that differences in deciding for and predicting the behavior of others is domain dependent, as the pattern is the reverse of what was found for low-impact relationship scenarios. Unfortunately, however, these results were based on only three of the money scenarios, as because of a programming error only three of the six money scenarios provided valid data. Thus these results should be viewed with caution, and we do not discuss them further. The individual difference measures were used to determine if different aspects of personality might influence decision making or prediction for the self and others differently. We found no significant interactions between any of the individual difference measures and condition, however. Note. Standard deviations are in parentheses. Sample size ranges from 25 to 29. For both the low-impact and high-impact scenarios, the minimum score is 0, indicating all risk-averse choices, and the maximum score is 9, indicating all risk-taking choices. 9To determine if aggregating over gender was reasonable, we conducted a 5 (condition) × 2 (gender) × 2 (impact level) repeated measures ANOVA on number of risk-taking responses. In keeping with previous studies, there was a strong main effect of gender, whereby men (M = 11.15) made more risk-taking decisions than did women (M = 8.75), F(1, 127) = 36.52, p .39). Therefore, collapsing across gender in Table 5 is appropriate, although all inferential analyses included gender in the model. 10Nonetheless, we repeated each of the planned comparisons reported in this and the subsequent section for high-impact scenarios to see if any differences would emerge in those scenarios. None of these contrasts were significant (all ps > .05).