Designing and Evaluating a Chatbot for Survivors of Image-Based Sexual Abuse

Title of the Paper

Design and Evaluation of a Chatbot Supporting Survivors of Image-Based Sexual Abuse

Paper Information

• Domain: Human-Computer Interaction (HCI), Social Technologies, Mental Health Support, Sexual Violence Prevention
• Keywords: Image-Based Sexual Abuse (IBSA), Revenge Porn, Chatbot, Informational Support, Emotional Support

Research Background and Problem Statement

• Key Issues or Challenges:

  1. Image-Based Sexual Abuse (IBSA) refers to the non-consensual sharing or creation of intimate images, which causes severe psychological, social, and physical issues for survivors, including depression, anxiety, and self-blame.
  2. Survivors often face social stigma, self-reproach, and a lack of legal protection and emotional support when seeking help.
  3. While online platforms can report and remove images, the technology is often difficult to navigate, and support systems are generally inaccessible and user-unfriendly. Emotional support is scarce, leading many survivors to remain silent.

• Significance of the Study:

  1. Surveys indicate that fewer than 10% of survivors successfully utilize legal or online tools to effectively address their situation.
  2. Providing timely, low-barrier, always-available support tools, such as chatbots, can address the current unavailability of assistance and alleviate survivors' emotional trauma.

• Research Motivation and Related Work:

  1. Numerous supportive chatbots have been developed to provide mental health support and legal assistance.
  2. This study aims to explore whether such technology can deliver informational and emotional support in the context of IBSA.
  3. The authors analyzed nine existing chatbots designed for survivors of sexual violence and collaborated with professional psychological counselors to derive design guidelines.

Solution

• Proposed Method or Solution:

  1. Developed a hybrid chatbot integrating rule-based modules and retrieval-based modules to provide informational guidance and emotional comfort for IBSA survivors.
  2. The chatbot includes seven modules: resource information, coping strategies, complaint letter writing guidance, IBSA knowledge, legal procedure explanation, survivor story sharing, and survivor rights.

• Innovations:

  1. The chatbot introduces technical innovations (using hybrid modules for informational guidance and emotional interaction) while offering highly professional support by integrating social policy resources.
  2. Unique modules tailored to IBSA survivors' needs were designed, such as complaint letter assistance and context-specific psychological support phrases.
  3. AI technology was utilized to optimize the chatbot experience through text analysis and personalized responses.

• Implementation Steps:

  1. Design Phase: Combined expert recommendations with seven functional modules, analyzed reference chatbots, and designed survivor-centric information structures.
  2. Development Phase:
     • Rule-based modules: Constructed decision tree structures based on procedural flows.
     • Retrieval-based modules: Used deep learning (BERT model) for user input classification and relevance-based response matching.
  3. Hybrid Implementation: Integrated rule-based and retrieval-based modules, allowing the system to switch to structured information guidance when retrieval-based responses fail to meet user needs.
  4. User Evaluation: Compared the chatbot's effectiveness in providing informational and emotional support against internet searches.

Research Outcomes

• Specific Results:

  1. Informational Support: Participants found the chatbot superior to internet searches, particularly in terms of information organization, timeliness, and content conciseness.
  2. Emotional Support: The comforting language provided by the chatbot effectively warmed participants, closely resembling human-provided emotional support.
  3. Methodological Advantages: The hybrid implementation allowed users to quickly query information via keywords while engaging in modular conversations to explore additional resources.

• Advantages Over Existing Solutions:

  1. Provides structured information, eliminating the need for survivors to independently devise coping strategies.
  2. The hybrid implementation addresses the limitations of purely rule-based chatbots in complex psychological support scenarios.
  3. Offers 24/7 real-time support without temporal or spatial constraints.

• Experiment or Evaluation Results:

  1. User Study Findings:
     • The chatbot significantly outperformed internet searches across 11 informational metrics, including "accessibility," "ease of use," and "accuracy," but slightly lagged in "trustworthiness" and "objectivity."
     • The chatbot surpassed internet searches across nine emotional dimensions, such as reducing self-blame and fostering acceptance.
  2. Suggestions for Improvement: Future designs should clearly specify the chatbot's development institution to enhance credibility. Emotional support modules require further optimization to better align with survivors' psychological states.

• Limitations and Future Directions:

  1. Current Limitations:
     • Experiments were conducted with non-survivors in hypothetical scenarios; future studies should directly involve actual survivors.
     • Scenarios were limited to a single type of IBSA case; subsequent research should expand to diverse contexts.
  2. Improvement Directions:
     • Fine-tune retrieval modules to better handle open-ended user inputs.
     • Provide more survivor-friendly emotional interactions, such as customized suggestions and real-life case connections.

Conclusion

• Through innovative practices in human-computer interaction, this study demonstrates the potential and feasibility of hybrid chatbots in delivering informational and emotional support. Its design principles and methods offer novel solutions for survivors of sexual violence with technological needs, while expanding application scenarios and design guidelines for the HCI research community.