Real-time assistance in suicide prevention helplines using a deep learning-based recommender system: A randomized...

Area of Science:

• Psychology
• Computer Science
• Public Health

Background:

• Suicide prevention helplines are critical for mental health support.
• Counselors face high-pressure situations requiring real-time decision-making.
• AI tools offer potential for augmenting human support in crisis interventions.

Purpose of the Study:

• To assess the effectiveness and usability of an AI-assisted tool for real-time counselor support in suicide prevention.
• To evaluate the impact of AI assistance on counselor self-efficacy and response times.
• To analyze the patterns and appropriateness of AI tool usage during helpline conversations.

Main Methods:

• A randomized controlled trial (RCT) comparing an AI tool group with a control group (care as usual).
• The AI tool utilized BERT sentence embeddings and cosine similarity for suggestion generation.
• Counselor self-efficacy was measured via questionnaires; tool usage was logged for frequency, duration, and content analysis.

Main Results:

• No significant difference in self-efficacy between AI and control groups (p=0.36).
• AI tool users exhibited marginally lower response times and increased usage in longer conversations.
• Appropriate AI tool use yielded usable suggestions in 83% of instances, but optimal timing was often missed.

Conclusions:

• AI-assisted tools show potential for supporting counselors in suicide prevention, offering usable suggestions.
• Challenges remain in optimal AI tool integration, including counselor proficiency and trust.
• Further research is needed to refine AI tools for effective real-time crisis intervention support.