Replication of changing hearts and minds? Why media messages designed to foster empathy often fail (Gubler et al., 2022)

This paper focuses on computational reproducibility and robustness replicability of Gubler et al.'s(2022) studies which examine the effect of media messages on empathic concern, dissonance, and out-group policy attitudes. The original paper tests four hypotheses using two online experiments with large samples from one US state (N1=5,800; N2=2,200). Regarding the first experiment, we successfully reproduced the effect that initial antipathy weakens the effect of humanizing treatment on empathic concern (H1). However, we show that the moderating effect is negligible and has little practical significance. Moreover, the individual effect estimates in our analyses slightly differed from the original paper due to different procedure of data cleaning and minor coding errors in the original paper. The most relevant difference was the opposite effect of gender than reported in the original paper. We also show that empathic concern might mediate the effect of humanizing treatment on attitudes toward immigrants (H3). The original study rejected the mediation hypothesis due to not finding a total effect of humanizing treatment on attitudes. In contrast, we found that humanization treatment has a positive indirect effect on attitudes through empathic concern. At the same time, it also has a direct negative effect on attitudes. For the second experiment (H1, H2a, H2b, H3), we attempted to reproduce the results using a different software. We partially succeeded once receiving support from the authors of the original study. We note throughout the report issues we have encountered.