Empathy-based counterspeech can reduce racist hate speech in a social media field experiment

Online hate speech poses significant risks to public discourse and the well-being of marginalized groups. While content moderation is a common response, it often faces criticism for potentially suppressing legitimate speech or merely displacing hate speech to other platforms. This study investigates whether 'counterspeech'—direct, non-suppressive responses to hate speech—can effectively persuade perpetrators to reduce their hostile behavior.

The researchers conducted a field experiment on Twitter, identifying 1,350 users who posted xenophobic or racist content. These users were randomly assigned to one of three intervention groups or a control group. The interventions were delivered by neutrally designed, human-controlled 'bot' accounts within 24 hours of the original post. The three strategies tested were:

• Empathy: Messages designed to humanize the victim and highlight the emotional impact of the hate speech.
• Warning of consequences: Messages reminding the user that their public posts are visible to friends and family.
• Humor: Messages using memes or humorous animal imagery to de-escalate the interaction.

The study measured outcomes including the retrospective deletion of the original tweet and the prospective creation of xenophobic content over a four-week follow-up period.

The results indicate that empathy-based counterspeech is the most effective of the three strategies. Users who received an empathy-based message were significantly more likely to delete their original xenophobic tweet and showed a measurable reduction in the volume of xenophobic content produced over the following month. In contrast, strategies relying on humor or warnings of consequences yielded no consistent, statistically significant effects. These findings suggest that appealing to the perpetrator's empathy is a promising, albeit modest, tool for mitigating online hostility.