How social relationships shape moral wrongness judgments

Abstract

Judgments of whether an action is morally wrong depend on who is involved and the nature of their relationship. But how, when, and why social relationships shape moral judgments is not well understood. We provide evidence to address these questions, measuring cooperative expectations and moral wrongness judgments in the context of common social relationships such as romantic partners, housemates, and siblings. In a pre-registered study of 423 U.S. participants nationally representative for age, race, and gender, we show that people normatively expect different relationships to serve cooperative functions of care, hierarchy, reciprocity, and mating to varying degrees. In a second pre-registered study of 1,320 U.S. participants, these relationship-specific cooperative expectations (i.e., relational norms) enable highly precise out-of-sample predictions about the perceived moral wrongness of actions in the context of particular relationships. In this work, we show that this 'relational norms' model better predicts patterns of moral wrongness judgments across relationships than alternative models based on genetic relatedness, social closeness, or interdependence, demonstrating how the perceived morality of actions depends not only on the actions themselves, but also on the relational context in which those actions occur.

Fig. 1. Kernel density plots of prescribed cooperative functions for 20 common relationship dyads.

Dots represent the population mean prescription for each cooperative function within each relationship, caps represent +/− one standard deviation. The height of the curve represents density: the likely proportions of scores (relative to each function) that fall within the given range along the x-axis. Source data are provided as a Source Data file.

Fig. 2. Relational norm profiles for a subset of 10 relationships.

Pink represents care, black represents hierarchy, green represents mating, blue represents reciprocity. The raw data (n = 423 independent ratings per function per relationship; total n = 16,920) are shown in individual dots; error bars represent the mean (dot) and + /− 1 SD (caps). Note: Mother/Father and under-18 child have been combined into a single plot. Plots for all 20 relationships are in Supplement Section 1.4.4. Source data are provided as a Source Data file.

Fig. 3. Hierarchical clustering of relationships.

Circular dendrogram visually representing the mean Kolmogorov-Smirnov (K-S) distance between relationships in four-dimensional relational norm space, clustered hierarchically according to the Voorhees method (a); relationships selected for Study 2 are highlighted in a darker shade. Radar plots derived from the hierarchical cluster model are depicted in the bottom half of the figure (b). The left panel shows the overlapping clusters; the right panel shows each cluster on its own set of axes. Source data are provided as a Source Data file.

Fig. 4. Characteristic function-weakening actions.

Heatmap showing mean ratings of judges (n = 15) of the extent to which each action characteristically would neglect or violate (weaken) the care, hierarchy, mating, and reciprocity functions, respectively, between any two people (i.e., not assuming the relationship between “Person A” and “Person B” should in fact serve any of those functions). These items were chosen as experimental stimuli from a much larger set by an algorithm using the judges’ ratings, where −100 represents the most characteristic function-weakening effect (see Methods). Darker shades represent more extreme ratings. Note: when rating actions on the hierarchy dimension, judges were asked to imagine that Person A was in a subordinate role, specifically; when rating actions on the care dimension, judges were asked to imagine that Person A was in a caregiving (as opposed to care-seeking) role, specifically. Source data are provided as a Source Data file.

Fig. 5. Moral wrongness judgments.

Sample 2 moral wrongness judgments for cooperative function violations in different relationships: kernel density plot of wrongness judgments (0 = not at all morally wrong, 100 = very morally wrong) concerning characteristic function-weakening actions for each of four dyadic cooperative functions across 10 relationships. Dot represents the mean, with 95% confidence intervals. Height of the curve represents density (see Fig. 1 for explanation). This experiment was conducted once, with all data shown here. Note that actions which characteristically weaken the mating function (e.g., refusing to have sex with someone) were judged closer to “not at all wrong” than “very wrong” for all dyads apart from the romantic partner relationship. Otherwise, the relative lack of visually dramatic differences in the shape of the moral wrongness judgment distributions between relationships likely can be explained by the mild or “everyday” nature of the function-weakening actions employed in this study (see Fig. 4). Such actions were deliberately chosen to contrast with the more extreme, unusual, or bizarre actions often studied in moral psychology; thus, the ability of our model to predict even subtle variance in moral wrongness judgments between relationships for common, non-extreme actions (see analysis below) can be seen as a strength. Source data are provided as a Source Data file.

Fig. 6. Correlation between dyad dissimilarity in relational norms and dissimilarity in moral judgments.

Scatterplot showing the predicted correlation in K-S distance between each pair of relationship dyads in relational norm space (x-axis) and the K-S distance between those same dyads in moral judgment space (y-axis). Spearman’s r = .43, p = .003. Note that the color of each relationship reflects the cluster in which it is located from Fig. 3. Source data are provided as a Source Data file.