Is there negative social influence? Disentangling effects of dissimilarity and disliking on opinion shifts

Empirical studies are inconclusive about the underlying mechanisms that shape the interrelated dynamics of opinions and interpersonal attraction. There is strong evidence that others whom are liked have a positive influence on opinions and similarities induce attraction (homophily). We know less about “negative” mechanisms concerning whether disliked others induce shifts away from consensus (negative influence), whether large differences (dissimilarity) generate distancing, and whether dissimilarities induce disliking (heterophobia). This study tests discriminating hypotheses about the presence of positive and negative mechanisms in controlled experiments involving dyadic interactions. Results confirm the presence of homophily, do not support the existence of negative social influence, and show a robust positive linear relationship between opinion distance and opinion shifts. This implies that contact might provide the largest push towards consensus in case of large initial differences.

💡 Research Summary

The paper investigates whether “negative social influence” – the idea that disliked others push us away from their opinions – actually operates in interpersonal opinion dynamics, and how it interacts with the well‑established positive mechanisms of homophily (similarity → attraction) and influence. The authors formulate four competing hypotheses: (1) liked partners exert a positive pull on one’s opinion (positive influence); (2) disliked partners cause a repulsive pull, moving one’s opinion away (negative influence); (3) larger opinion differences generate heterophobia, reducing attraction (dissimilarity → dislike); and (4) opinion distance itself predicts the magnitude of opinion change, possibly in a linear fashion.

To test these hypotheses, the authors conduct controlled dyadic experiments in a laboratory setting. Participants first express their stance on a continuous issue (0–100 scale), then view a randomly assigned partner’s stance, rate their attraction to the partner on a 1–7 Likert scale, engage in a brief structured conversation, and finally re‑report their own stance. The sample consists of 200 dyads (400 individuals). The key variables are initial opinion distance, initial attraction, and post‑interaction opinion shift. The authors analyse the data with mixed‑effects regression models, treating participants and dyads as random effects to account for repeated measures and individual heterogeneity.

Results show a robust homophily effect: smaller opinion distances are associated with higher attraction (p < .001). However, the data provide no support for a negative influence effect; even when attraction is low, participants still move their opinions toward the partner’s position, and there is no systematic repulsion. Likewise, the hypothesised heterophobia effect is absent – opinion distance does not predict lower attraction. The most striking finding is a strong positive linear relationship between opinion distance and the size of the opinion shift (p < .001). In other words, the larger the initial disagreement, the larger the adjustment toward the partner’s view. This “distance‑driven convergence” pattern suggests that interpersonal contact is most potent for reducing disagreement when initial differences are large.

The authors interpret these findings as a challenge to models that embed a direct “dislike → repulsion” pathway. Instead, they argue that people appear to be motivated to reduce disagreement regardless of affective valence, and that the magnitude of the pull is proportional to how far apart the initial positions are. The paper discusses methodological strengths (random assignment, precise continuous measures, mixed‑effects modeling) and acknowledges limitations: the artificial laboratory context, a sample dominated by university students, and the absence of real‑world stakes (e.g., political or moral issues). Future work is suggested to replicate the design across cultures, age groups, and online environments, and to incorporate additional variables such as perceived expertise or group identity.

Practically, the results imply that fostering dialogue between groups with divergent views can be an effective consensus‑building tool, especially when the gaps are wide. Policymakers, organizational leaders, and conflict‑resolution practitioners might therefore prioritize structured interactions across ideological divides, rather than assuming that large differences inevitably breed hostility or that disliked interlocutors will push opinions further apart. In sum, the study confirms homophily, refutes the existence of negative social influence under the tested conditions, and uncovers a linear, distance‑driven mechanism that pushes opinions toward convergence.