Online perspective-taking experiments have demonstrated great potential in reducing prejudice towards disadvantaged groups, but had no meaningful causal effect on social welfare attitudes

Bor, Alexander, and Gábor Simonovits. 2020. “Empathy, Deservingness, and Preferences for Welfare Assistance: A Large-scale Online Perspective-taking Experiment.” PsyArXiv. January 29. doi:10.31234/osf.io/d4sm9

Abstract: Online perspective-taking experiments have demonstrated great potential in reducing prejudice towards disadvantaged groups such as refugees or the Roma. These studies trigger the psychological process of empathy and evoke feelings of compassion. Meanwhile, a growing literature argues that compassion towards the poor is an important predictor of support for social welfare. This paper bridges these two literatures and predicts that perspective-taking with the poor could increase support for welfare assistance. This hypothesis is tested with a pre-registered experiment conducted on a large and diverse online sample of US citizens (N=3,431). Our results suggest that participants engaged with the perspective-taking exercise, wrote eloquent, often emotional essays. Nevertheless, perspective-taking had no meaningful causal effect on social welfare attitudes; we can confidently rule out effects exceeding 2 points on a 100 points scale. These results cast serious doubt on perspective-taking as a viable online tool to create compassion towards the poor.


Discussion

In this paper, we have tested whether perspective-taking is a viable tool for increasing support for welfare redistribution. Relying on an original, carefully designed, well-powered, and pre-registered survey experiment elded to a representative sample of US citizens, we found that it is not. Similarly to successful interventions, we tested the impact of a particular stimulus, describing the experiences of a single target and emphasizing a particular set of challenges that poor people in the US face (unemployment, health problems, housing problems, single parenthood).
Thus, our conclusions about the possible e ectiveness of perspective-taking intervention are necessarily limited: We have no way of knowing if large or even small changes in the stimulus used here could have led to a more e ective intervention.
This leads to the question of the extent to which these null ndings advance our understanding of either the class of interventions or the substantive target attitude that we study.
This issue should be understood in the broader context of how the published experimental literature characterizes the e ect of a di erent class of interventions. There is ample evidence that published research over-represents successful interventions compared to the universe of social science experiments (Franco, Malhotra and Simonovits 2014; 2016). For a more complete understanding of how a given class of interventions – such as perspective-taking – works, one also needs to consider unsuccessful examples.
That said, it is also important to emphasize why we think that our null results are surprising. First, our experimental design relied on a heavy dose of deservingness cues, which, according to previous research, has a large and sometimes long-lasting e ect on support for redistribution. We expected the perspective taking exercise to amplify the e ects of these deservingness cues but found that it nulliffed it.
Second, our findings are surprising considering a growing line of research employing perspective taking to reduce prejudice against various groups from refugees to transgender individuals.
Our results suggest that prejudice against the poor and attitudes towards government help for the poor may be more di cult to shape than attitudes towards these other marginalized groups.
Third, our experimental design likely constitutes a liberal test of our hypothesis. Besides relying heavily on deservingness cues, we measure the dependent variable with a composite index of ten items after a distractor task lasting a few minutes. For this reason, the experiment should be able to pick up even small, fleeting effects. Finally, it is noteworthy that the analysis of the essays reveals that participants have been very attentive and engaged in the exercise. We have no reason to believe that if we had conducted our experiment in the lab, we would see different results.
At the same time, the literature also o ers some explanations for our failure to bring about attitude change using our perspective-taking intervention. On the one hand, our treatment might have proven too weak in the sense that even though subjects felt empathy towards the individual depicted in the vignette, these emotions did not spill over to people in need in general, perhaps because subjects viewed the vignette as an \exception" to some deeply held stereotypes about poor people.
On the other hand, intense exposure to a story about a person in need might have led to emotional reactions moving counter to our hypothesized e ect. For instance, as argued by Sands (2017) exposure to poverty might have provoked anxiety in subjects about their own relative status, suppressing their support for policies helping others. Similarly, as pointed out by Simas, Clifford and Kirkland (2020), heightened empathy might exacerbate in-group bias, leading to hostile attitudes towards members of an out-group.