Effectiveness of Hybrid AI and Human Suicide Detection Within Digital Peer Support (Preprint)

BACKGROUND Suicidality rates continue to rise, while mental health services face obstacles of access, availability, and affordability. Digital peer support (DPS) shows promise in bridging these gaps and serving as an early step in identifying and addressing suicidal ideation (SI). OBJECTIVE This study examined: (1) the effectiveness of a hybrid solution combining a proprietary SI detection AI model and real-time human moderation within DPS, (2) demographic distribution of SI, (3) speed of active SI users’ referral to crisis care, (4) linguistic differences between communication of passive and active SI (5) sentiment changes among DPS users who express passive SI, and (6) effects of peer SI disclosure (passive and active) on users without SI. METHODS We retrospectively analyzed 169,181 live chat transcripts encompassing 449,946 user visits (January - December 2024) among peers from a DPS provider, Supportiv. Passive and active SI were identified using a hybrid AI model plus human moderator solution with post-hoc LLM verification. Sentiment analysis and ANCOVA compared changes in despair, loneliness, helplessness, despair, and optimism across three propensity-matched user groups: passive SI users, non-SI users exposed to SI peer content, and non-SI users not exposed to SI. RESULTS SI was present in 3.19% of the live chats. The AI model flagged SI faster than humans (in 77.52% passive, 81.26% active cases), with 90.3% agreement. Moderators followed up on average 71.3 seconds after AI alerts and referred 5,472 active SI users (1.21%) to higher acuity crisis care. All users, including those with SI and those exposed to peer SI, significantly benefited from DPS. Sentiment scores were reduced by up to 29.3% for depression, 26.8% for loneliness, 25.3% for despair, and 22.3% for helplessness, whereas optimism increased by up to 40.4%. CONCLUSIONS DPS with AI integration offers scalable support for high-risk populations. The proprietary SI detection AI model accurately detects suicidality, allowing for human-moderated DPS to improve the mental well-being for users with and without SI, while maintaining the safety and emotional stability of peers exposed to SI.