Understanding and detecting behaviours prior to a suicide attempt: A mixed-methods study

Area of Science:

• Public Health
• Computer Science
• Psychiatry

Background:

• Observable behaviors precede suicide attempts in public spaces.
• Current monitoring limitations hinder timely interventions.
• Need for continuous monitoring systems for suicide prevention.

Purpose of the Study:

• Examine the acceptability and feasibility of automated computer systems for crisis behavior detection.
• Assess public perceptions of using CCTV and AI for suicide prevention.
• Develop and evaluate a computer vision algorithm for identifying pre-suicidal behaviors.

Main Methods:

• Mixed-methods study combining surveys and video analysis.
• Large-scale acceptability survey on CCTV and AI for suicide prevention.
• Manual analysis of CCTV footage to identify crisis behaviors.
• Development and evaluation of a computer vision algorithm for behavior recognition.

Main Results:

• Public attitudes towards using CCTV and AI for suicide prevention are positive.
• Identifiable crisis behaviors include repetitive pacing and extended stays in public locations.
• The computer vision algorithm achieved 80% sensitivity and 90% specificity in detecting crisis behaviors.

Conclusions:

• Automated behavior recognition using computer vision is feasible for suicide prevention.
• Community acceptance is high for AI-driven suicide prevention technologies.
• This approach may offer a viable method for initiating human contact during crises.