Echoes of Norms: Investigating Counterspeech Bots’ Influence on Bystanders in Online Communities

Paper Title

Echoes of Norms: Investigating Counterspeech Bots’ Influence on Bystanders in Online Communities

Publication Info

• Topic area: Counterspeech bots and their impact on bystanders in online hate speech contexts.
• Keywords: Counterspeech, hate speech, online communities, bystanders, chatbot design, cognitive strategies, affective strategies, tone, sentence type.

Background and Problem

• Problem / challenge: Existing research on counterspeech chatbots has focused on curbing hate speakers and supporting targets, but their influence on bystanders—arguably the most critical group in shaping community norms—remains unclear.
• Significance: Bystanders form the "silent majority" in online communities and significantly influence prevailing attitudes and norms. Understanding how counterspeech affects them is crucial for fostering healthier online environments.
• Motivation and related work: Prior studies have highlighted the promise of counterspeech in shaping norms through contagion, emotion, and identity influence. However, systematic empirical evidence on chatbot-mediated counterspeech’s impact on bystanders is limited, leaving a gap in understanding its broader social influence.

Solution

• Proposed approach: Development of a counterspeech strategy framework and creation of Civilbot, a prototype chatbot capable of generating diverse counterspeech responses based on sentence type, tone, and strategic intent.
• Novelty:
  1. First study to examine mechanisms through which counterspeech chatbots influence bystanders in online hate incidents.
  2. Development of a unified experimental framework integrating sentence type, tone, and strategic intent into eight distinct counterspeech strategies.
  3. Provision of design insights for counterspeech chatbots that explicitly consider the role of bystanders in shaping responses to hate speech.
• Procedure and key techniques:
  • Literature review to synthesize counterspeech strategies and develop a unified framework.
  • Construction of Civilbot, capable of generating counterspeech responses across eight strategy combinations.
  • Mixed-methods within-subjects experiment with 52 participants to evaluate Civilbot’s influence on perceived counterspeech quality, subjective acceptance, and behavioral tendencies.

Results

• Concrete findings:
  • Civilbot was perceived as credible (3.33/5) and norm-affirming but scored lower on persuasiveness metrics like "strong reason" (2.84/5).
  • Positive-tone cognitive strategies significantly outperformed affective ones in perceived quality (F = 24.20, p < 0.001, Cohen's f = 0.34).
  • Behavioral influence was subtle, including guidance, substitution, negative modeling, and reverse motivation effects.
• Advantage over baselines: Civilbot’s structured counterspeech framework provided nuanced insights into strategy effectiveness, outperforming generic counterspeech approaches in fostering community-level dialogue and critical reflection.
• Experiments / evaluation:
  • Participants selected eight hate speech topics and experienced counterspeech responses across all strategy types.
  • Metrics included perceived counterspeech quality, subjective acceptance, and pre–post changes in behavioral tendencies.
  • Data analysis combined quantitative methods (ANOVA, t-tests) with qualitative thematic analysis of participant feedback.
• Limitations and future work:
  • Short-term lab study may suppress natural behaviors; future longitudinal field studies are needed.
  • Context-sensitive adaptive mechanisms for counterspeech strategies should be explored.
  • Cross-cultural studies are required to assess generalizability beyond the Chinese context.
  • Real-time detection and generation capabilities should be integrated for practical deployment.

Summary

This study investigates the influence of counterspeech chatbots on bystanders in online hate speech contexts, focusing on Civilbot, a prototype chatbot designed to deliver context-aware counterspeech across eight strategies. Findings reveal that Civilbot is perceived as credible and norm-affirming but constrained by shallow reasoning. Cognitive strategies paired with positive tone were most effective, while affective strategies showed situational value. Behavioral effects were nuanced, ranging from guidance to reverse motivation. Civilbot also contributed to shaping community climate by providing information, cooling emotions, and encouraging critical reflection. These insights inform the design of counterspeech bots that adapt strategies to context, integrate credible evidence, and expand beyond text-based modalities to foster healthier online environments.