Designing a Conversational Agent for Sexual Assault Survivors: Defining Burden of Self-Disclosure and Envisioning Survivor-Centered Solutions

Title of the Paper

Designing Conversational Agents for Survivors of Sexual Assault: Defining the Burden of Self-Disclosure and Exploring Survivor-Centered Solutions

Paper Information

• Subject Area: Applications of Human-Computer Interaction (HCI) and Computer-Supported Cooperative Work (CSCW) in Mental Health and Support for Survivors of Sexual Violence
• Keywords: Conversational Agent (CA), AI Chatbot, Burden of Self-Disclosure, Sexual Assault Survivors, Survivor Support

Research Background and Problem Statement

• Identified Problems or Challenges:

  • Survivors of sexual assault often hesitate to share their experiences with others or report incidents due to psychological, social, and cultural reasons, leaving them without necessary support.
  • The phenomenon of "secondary victimization" during self-disclosure (e.g., blame, shame) exacerbates survivors' psychological burdens.
  • Traditional human-computer interaction systems often overlook users' psychological burdens and subjective experiences when designing technologies for survivors.
  • Existing chatbot products demonstrate design deficiencies when addressing sensitive topics, potentially leading to misunderstandings or harm.

• Why It Matters:

  • Providing safe, autonomous, and non-judgmental tools for survivors of sexual assault, particularly those under significant psychological and social pressure, is crucial.
  • Designing a conversational agent that reduces psychological burdens could increase the likelihood of reporting sexual offenses or seeking help while lowering barriers in the reporting process.

• Research Motivation:

  • Although existing chatbots have shown potential in supporting mental health, the specific needs of survivors of sexual violence remain inadequately addressed.
  • Exploring how technology can reduce the burden of self-disclosure and reporting for survivors, thereby promoting their well-being and quality of life.

Solution

• Proposed Methods or Solutions:

  • Developed a conversational agent prototype (NamuBot) aimed at providing informational, functional, and emotional support to alleviate the burdens encountered during the process of reporting sexual crimes.
  • Collaborated with professionals related to sexual assault (e.g., police officers, counselors) during the early design phase to define needs and conceptualize the design.
  • Incorporated user needs and feedback through participatory design involving survivors, police officers, and counselors in later stages.

• Innovative Aspects:

  • Engaged survivors, police officers, and counselors as key stakeholders to co-design a system focused on identifying and alleviating the psychological and practical burdens of self-disclosure for survivors.
  • The system's functional design prioritized survivors' needs, including privacy protection, enhanced security, automatic data saving, and a discreet user interface.
  • Introduced self-compassion features and safety mechanisms, such as multi-layer authentication and pseudonymized conversation content, to enhance usage security.

• Implementation Steps and Key Technologies:

  • Conducted needs assessments during the initial design phase, including thematic analysis and stakeholder needs analysis.
  • Developed a hybrid logic conversational agent based on rule-based and deep learning approaches, enabling survivors to interact via text or voice.
  • Completed iterative design through prototype testing and focus groups (including survivors and professionals), resulting in user-driven improvement strategies and solutions.

Research Outcomes

• Specific Findings:

  • Identified and categorized nine types of burdens faced by survivors during self-disclosure, including time burden, financial burden, privacy and security burden, social burden, and data management burden.
  • Found that most survivors preferred interacting with a conversational agent over human-to-human interaction (HHI), with significantly reduced psychological burdens when engaging with NamuBot.
  • Proposed and designed seven features to address burdens introduced by conversational agents themselves, such as multi-layer authentication, data backup and auto-lock mechanisms, and disguised user interfaces.

• Advantages Compared to Existing Solutions:

  • Delivered a highly personalized, survivor-friendly technological design that meets survivors' high demands for privacy and security.
  • Demonstrated superior performance in operational convenience, emotional support, and anonymity, expanding the technology's applicability to marginalized survivor groups.
  • Experimental results showed that NamuBot better addressed core survivor needs when disclosing sensitive information compared to existing technologies.

• Limitations and Future Directions:

  • The study primarily focused on user samples within the cultural context of South Korea, which may limit its direct applicability to other regions.
  • Excluded research on minor survivors and male users, necessitating future expansion of the study population.
  • Future work should further validate the feasibility and scalability of the developed system, particularly in the context of global cultural and legal variations.

Conclusion

This study developed a conversational agent prototype (NamuBot) specifically designed for survivors of sexual violence through collaboration with survivors and professionals, offering new design insights and reference standards for the application of related technologies in psychological and social support. The research highlighted how technological design can alleviate psychological and practical burdens in high-stress, high-risk environments. This study fills a critical gap in current technological support tools for sensitive issues and lays a solid foundation for future related research.