Collaborative Upstanding: Exploring Conversational Strategies for Cyberbullying Upstanding Education

Paper Title

Collaborative Upstanding: Exploring Conversational Strategies for Cyberbullying Upstanding Education

Publication Info

• Topic area: Human-AI collaboration for cyberbullying intervention and upstanding education.
• Keywords: Cyberbullying, bystander intervention, upstanding, chatbots, human-AI collaboration, conversational AI, educational technology, social computing, situational barriers, public intervention.

Background and Problem

• Problem / challenge: Cyberbullying persists as a significant issue on social media, and while bystander intervention (upstanding) is effective, bystanders face barriers such as self-efficacy doubts, lack of knowledge on how to intervene, and fear of repercussions. Existing educational frameworks and systems focus on single stages of intervention and lack in-situ, interactive solutions.
• Significance: Encouraging upstanding can mitigate cyberbullying's harmful effects on victims and foster a supportive online environment. Addressing barriers to upstanding is critical for empowering bystanders to act effectively.
• Motivation and related work: Prior research has explored bystander intervention models, educational frameworks like STAC, and the potential of AI for role-playing and training. However, no prior work has implemented and evaluated chatbots for real-time bystander training. This paper builds on these gaps to propose a collaborative upstanding approach using conversational AI.

Solution

• Proposed approach: ConCUR (CONversational Collaborative Upstanding in Real-time), a chatbot system designed to guide bystanders through the upstanding process by focusing on co-authoring upstanding messages and addressing barriers flexibly and iteratively.
• Novelty:
  1. Introduction of a collaborative upstanding approach where bystanders and AI co-pilots work together to address barriers.
  2. Implementation of a how-to-first conversational structure to prioritize actionable guidance.
  3. Identification of effective conversational strategies for overcoming upstanding barriers.
  4. Empirical evaluation of human-AI collaborative upstanding in lab settings.
• Procedure and key techniques:
  • Study 1: Paired role-play study (n=24) to identify barriers and conversational strategies in human-human collaborative upstanding.
  • Study 2: Implementation and evaluation of ConCUR (n=20), focusing on how-to-first interactions and their impact on motivation and self-efficacy.
  • Analysis of conversational patterns, barriers, and strategies through qualitative and thematic coding.

Results

• Concrete findings:
  • Study 1: Identified key barriers (e.g., how-to challenges, self-efficacy doubts, fear of repercussions) and effective conversational strategies (e.g., probing questions, aiding situational interpretation, cognitive reframing).
  • Study 2: 90% of participants successfully posted public comments in a lab setting, with many reporting increased motivation and confidence for future upstanding.
• Advantage over baselines:
  • The how-to-first approach allowed simultaneous addressing of multiple barriers, unlike traditional sequential models.
  • ConCUR facilitated practical skills-building and increased participants' self-efficacy and willingness to intervene.
• Experiments / evaluation:
  • Study 1: Paired role-play with 24 participants aged 18–21, analyzing chat transcripts and interviews to identify barriers and strategies.
  • Study 2: Lab study with 20 participants aged 18–22, evaluating ConCUR's effectiveness in promoting public active upstanding.
  • Metrics included task completion rates, participant feedback, and qualitative analysis of chat interactions.
• Limitations and future work:
  • Lab settings may not fully represent real-life behaviors; long-term effects were not measured.
  • Limited evaluation of upstanding message quality.
  • Future work should explore real-world deployment, multi-outcome intervention models, and responses to evolving online harm dynamics.

Summary

This paper introduces ConCUR, a chatbot system designed to promote cyberbullying upstanding through collaborative, how-to-first conversational interactions. Two studies identified key barriers to upstanding, such as how-to challenges and self-efficacy doubts, and demonstrated the effectiveness of conversational strategies in overcoming these barriers. ConCUR increased participants' motivation and confidence to intervene, with 90% successfully posting public comments in a lab setting. The findings suggest that flexible, iterative, and in-situ conversational AI systems can play a significant role in upstanding education. Future research should explore real-world applications, long-term impacts, and adaptations to evolving online harm contexts.