Position paper on symbiotic intelligence in healthcare: Can AI help us better understand suicidal behavior and prevent suicide?

Suicide is the second leading cause of death among young people aged 10-24 worldwide, yet identifying individuals at risk remains a major challenge. In Norway, suicide rates reached their highest level in 25 years in 2024, underscoring persistent knowledge gaps in national prevention efforts and the need for innovative, interdisciplinary, and supplementary approaches. There are no simple solutions to mental ill health or suicide, and no single factor can adequately explain such complex phenomena; however, it remains crucial to examine how interacting risk factors may increase vulnerability and whether artificial intelligence can help identify emerging risk windows earlier, thereby complementing conventional clinical approaches in this field. The current evidence base suggests a need for increased vigilance in several areas, particularly regarding Gen Z's digital lifestyle, exposure to shock-like societal events, and patterns of alcohol consumption, as these factors may interact in ways that elevate suicide risk. The rapid growth of social media use over the past decade have given both a Werther- and Papageno-effects, raising questions about both the positive effects of social media use and also whether certain patterns of social media engagement may contribute to suicide risk among vulnerable groups. Emerging evidence indicates that shock events in society and extreme media exposure may affect vulnerable individuals indirectly and without conscious awareness, thereby increasing short-term suicide risk. Research from Norway and other countries shows that traumatic events may trigger acute spikes in suicides, influence perinatal outcomes, and affect population-level health indicators such as sex ratios and infant mortality. These "triple-hit" patterns suggest that indirect exposure through media may be more consequential than previously assumed. Additionally, alcohol use may play a significant role in short-term risk escalation by increasing impulsivity, reducing cognitive control, and intensifying emotional distress. International studies show that alcohol use is significantly associated with increased suicidality, and in Norway approximately four in ten individuals who die by suicide have alcohol in their bloodstream at the time of death. Early-warning systems could thus benefit from integrating alcohol-related indicators. This position paper argues that three opportunities are particularly salient. First, the majority of primary studies within this area rely on conventional research designs and analytical approaches, with limited use of artificial intelligence-supported methods that could potentially enhance measurement precision, validity, and reliability in the analysis of complex digital behaviors. For example, AI-based linguistic analysis of social media content may help detect short-term "risk windows" associated with psychological distress, depression, and suicidality. Second, improved access to anonymized, high-quality platform data from technology companies could strengthen population-level monitoring and research. Third, actigraphy and AI integrated with wearable sensors and brief daily ecological momentary assessments (EMA) may capture subtle fluctuations in sleep, stress, heart rate, and activity-patterns that often precede clinical deterioration but may go unnoticed by patients, families, and clinicians. While international studies suggest that AI can enhance short-term risk detection, such systems must neither replace human contact nor override core principles of privacy, consent, and autonomy. Rather, AI should function as a complementary, real-time alert layer (symbiotic intelligence) capable of informing timely and tailored interventions within existing health services. Given current knowledge gaps and the rising impact of shock-related stressors, health authorities should consider piloting AI-supported early-warning systems that are tightly embedded in clinical pathways, e.g., through a new conceptual model presented in this paper. AI alone will not save lives, but small, ethically grounded steps may help identify individuals in rapidly escalating distress before it is too late.