Interventions for Softening Can Lead to Hardening of Opinions: Evidence from a Randomized Controlled Trial

Andreas Spitz, Ahmad Abu-Akel, Robert West

Introduction

In an era of unprecedented online polarization, the prevailing narrative suggests that "echo chambers" and "filter bubbles" insulate individuals from opposing viewpoints, thereby entrenching their existing beliefs. However, previous research has shown that simply exposing users to dissenting opinions can often backfire, increasing hostility due to confirmation bias. This study investigates a potential remedy: the messenger. By leveraging the "pseudo-personal" or parasocial bonds people form with celebrities, the researchers explored whether a liked and trusted source could "soften" extreme views on controversial topics.

Methodology

The researchers conducted a survey-based Randomized Controlled Trial (RCT) with 379 US-based participants recruited via Amazon Mechanical Turk. The experimental design consisted of three distinct phases:

  1. Screening Survey: Participants were evaluated on their stances regarding four polarized topics—Climate Change, Vaccination, Immigration, and Abortion—and their attitudes toward a pool of 25 well-known celebrities (e.g., Tom Hanks, Ariana Grande).
  2. Randomized Assignment: Participants were exposed to polarized quotations (extracted from a corpus of 129 million news articles) that were randomly attributed to different spokespersons. This created four experimental conditions: Disagree/Like (opposing view from a liked celebrity), Disagree/Dislike (opposing view from a disliked celebrity), Disagree/Expert (opposing view from a fictitious expert), and Agree/Dislike (supporting view from a disliked celebrity).
  3. Measurement: Post-treatment changes were measured using a 7-point Likert scale for opinions and an "empathy-for-pain" test to gauge emotional shifts toward the spokesperson. Additionally, linguistic analysis (LIWC) was performed on written summaries provided by the participants.
Figure 1: Overview of the experimental design, including screening, randomized assignment, and the main survey structure.

Key Results and Analysis

Contrary to the hypothesis that liked messengers would bridge ideological divides, the study found no evidence of depolarization in any of the experimental conditions. Instead, the data revealed significant "backfire effects" where participants' original stances became even more entrenched.

Opinion Hardening

The most pronounced hardening of opinions occurred in two specific scenarios. First, the Disagree/Expert condition (Cohen’s $d = 0.24$) triggered a defensive response. The authors suggest that because an expert's view carries the weight of scientific authority, it may be perceived as a more direct threat to a person's worldview than a celebrity's opinion. Second, the Agree/Dislike condition (Cohen’s $d = 0.25$) led to a "justification effect." Participants seemingly reasoned that if even a person they disliked agreed with them, their original position must be undeniably correct.

Punitive Empathy Loss

Disagreement led to a universal drop in empathy toward the spokesperson, regardless of their prior status. Notably, the largest drop in empathy occurred in the Disagree/Like condition (Cohen’s $d = 0.44$). This suggests that when a liked celebrity voices an opposing view, participants perceive it as a "breach of trust," leading to a punitive emotional withdrawal.

Figure 3: Mean changes in standardized opinion (top) and empathy-for-pain (bottom) across different topics and experimental conditions.

Linguistic Shifts

Linguistic analysis mirrored these psychological shifts. In the Agree/Dislike condition, there was a significant increase in positive affective language ($p < 0.001$), reflecting the self-validation of being "right." Conversely, disagreeing with a liked celebrity was associated with an increase in the use of "anger" words, further highlighting the emotional friction caused by ideological dissent from a trusted source.

Discussion and Limitations

The findings suggest that the source of a message is not a simple "silver bullet" for reducing polarization. In fact, the very mechanisms intended to soften opinions—expertise and personal affinity—can trigger defensive hardening or emotional punishment when they clash with a person's core beliefs.

Several limitations should be noted. The participant pool was largely liberal-leaning, meaning the results primarily reflect how this demographic reacts to conservative-leaning prompts. Furthermore, the study measured immediate reactions to single interactions; long-term, repeated exposure might yield different results. Future research could investigate "triadic" interactions, where a trusted third party mediates the disagreement between a user and a spokesperson.

Conclusion

This RCT provides sobering evidence that traditional outreach strategies using liked celebrities or experts may be counterproductive in highly polarized environments. Rather than softening views, these interventions can lead to a hardening of opinions and a significant erosion of empathy, suggesting that the path to depolarization requires more complex strategies than simply changing the messenger.

Summary updated

Abstract

Motivated by the goal of designing interventions for softening polarized opinions on the Web, and building on results from psychology, we hypothesized that people would be moved more easily towards opposing opinions when the latter were voiced by a celebrity they like, rather than by a celebrity they dislike. We tested this hypothesis in a survey-based randomized controlled trial in which we exposed respondents to opinions that were randomly assigned to one of four spokespersons each: a disagreeing but liked celebrity, a disagreeing and disliked celebrity, a disagreeing expert, and an agreeing but disliked celebrity. After the treatment, we measured changes in the respondents' opinions, empathy towards the spokespersons, and use of affective language.Unlike hypothesized, no softening of opinions was observed regardless of the respondents' attitudes towards the celebrity. Instead, we found strong evidence of a hardening of pre-treatment opinions when a disagreeing opinion was attributed to an expert or when an agreeing opinion was attributed to a disliked celebrity. We also observed a pronounced reduction in empathy for disagreeing spokespersons, indicating a punitive response. The only celebrity for whom, on average, empathy remained unchanged was the one who agreed, even though they were disliked. Our results could be explained as a reaction to violated expectations towards experts and as a perceived breach of trust by liked celebrities. They confirm that naive strategies at mediation may not yield intended results, and how difficult it is to depolarize---and how easy it is to further polarize or provoke emotional responses.
BibTeX
@inproceedings{spitz2021interventions,
  author = {Andreas Spitz and Ahmad Abu-Akel and Robert West},
  title = {{Interventions for Softening Can Lead to Hardening of Opinions: Evidence from a Randomized Controlled Trial}},
  booktitle = {The Web Conference},
  year = {2021},
  doi = {10.1145/3442381.3450019},
}
Figure