"If This Person is Suicidal, What Do I Do?": Designing Computational Approaches to Help Online Volunteers Respond to Suicidality

Title of the Paper

“If This Person is Suicidal, What Do I Do?”: Designing Computational Approaches to Help Online Volunteers Respond to Suicidality

Paper Information

• Research Area: Human-Computer Interaction (HCI), Online Mental Health Support, AI-Assisted Technologies
• Keywords: Suicide Risk, Mental Health, Online Communities, Social Support, Artificial Intelligence, Content Moderation

Research Background and Problem

• Problem or Challenge:

  • Although many online platforms provide spaces for mental health support, strict restrictions on discussing suicidal behaviors pose challenges for volunteers trying to help individuals with suicidal tendencies. Volunteers face complex emotions and behaviors but lack specialized support.
  • Current technologies, such as suicide detection models and chatbots, are primarily used for content moderation or as substitutes for human support, rather than genuinely assisting volunteers.

• Significance:

  • According to data from the Centers for Disease Control and Prevention (CDC), in 2021 alone, approximately 12 million U.S. adults had suicidal thoughts, 1.7 million attempted suicide, and nearly 50,000 died by suicide.
  • Online platforms are crucial resources for many seeking help, especially for those unwilling or unable to access medical services.

• Research Motivation and Related Work:

  • Existing studies mostly focus on the needs of those seeking support, with insufficient research on the challenges and needs of volunteers providing support.
  • Suicide prevention technologies often emphasize automated interventions, with limited evaluation of their role in assisting volunteers during practical operations.

Solution

• Proposed Method or Solution:

  • Conduct a two-phase study to explore the practices and challenges of volunteers in identifying and responding to suicide risks, and evaluate how technological designs (e.g., AI) can provide support.
  • Design technology solutions based on suicide detection and collect volunteer feedback using the Speed Dating method to analyze effective design directions.

• Innovative Aspects:

  • Focus on optimizing human care infrastructure rather than merely developing technologies that replace human efforts.
  • In-depth exploration of how volunteers work under policies restricting discussions of suicide, and experimentation with collaborative technology designs to support volunteers.

• Implementation Steps and Key Technologies:

  • Phase 1: Conduct interviews and simulated chats to explore how volunteers assess the severity of suicide risk and their response strategies.
  • Phase 2: Propose and evaluate 18 technology design concepts, including emotional support tools, training resources, and suicide detection models.
  • Use participatory design (Speed Dating) to gather authentic feedback from volunteers and adjust solution directions accordingly.

Research Outcomes

• Specific Findings:

  • Online volunteers can accurately distinguish the severity of suicide risks but lack adequate guidance and support.
  • Suicide detection models help reduce volunteer stress but their automated decisions may not align with volunteers' actual needs, potentially causing privacy concerns and unintended harm.
  • Training chatbots based on simulators are a well-received solution, helping volunteers better prepare for conversations with individuals exhibiting suicidal tendencies.

• Comparative Advantages:

  • The proposed technology design solutions are more practical and context-sensitive in supporting volunteer practices compared to purely automated or content moderation technologies.
  • Encourages the establishment of non-medicalized support spaces, providing a safe environment for individuals with suicidal tendencies who reject traditional medical support for various reasons.

• Experimental or Evaluation Results:

  • Volunteers found AI simulators for training feasible and beneficial, reducing their stress during initial interactions with real support seekers.
  • Various emotional support tools effectively eased volunteer burdens, though concerns about privacy protection and algorithmic boundaries remain.

• Limitations and Future Directions:

  • Limitations:

    • Small sample size may lead to selection bias; participants with high interest in suicide research might not represent the general volunteer population.
    • Design concepts were proposed by the research team; adopting open co-creation design methods could yield richer participant perspectives.

  • Future Directions:

    • Investigate more diverse volunteer groups, including those from different countries and cultural backgrounds.
    • Explore ways to balance technological assistance with human autonomy to avoid potential privacy issues and inappropriate interventions for supporters or seekers.
    • Continue optimizing technologies for online suicide risk support, particularly in addressing the mental health burden on non-professional volunteers.