Machine Learning Algorithms Struggle to Predict Suicide Risk
A systematic review and meta-analysis of 53 studies reveals that machine learning algorithms have limited ability to reliably identify individuals at high risk for suicide or self-harm.
A new study published in PLOS Medicine (2025; Doi: 10.1371/Journal.pmed.1004581) indicates that machine learning algorithms are “hardly able to reliably identify people with high risk for suicide or self-harm.” The findings come from a systematic review and meta-analysis encompassing 53 studies.
For the past 50 years, numerous scales designed for risk assessment have been developed with the aim of identifying individuals with an elevated risk of suicide or self-harm. Though, these scales have demonstrated limited accuracy in predicting such risks.
“Melbourne-Machine-Learning algorithms are hardly able to reliably identify people with high risk for suicide or self-harm.”
Limitations of Current Risk Assessment Tools
The study highlights the ongoing challenge of accurately predicting suicide risk, despite the growth of various assessment tools. The reliance on machine learning, while promising, has not yet yielded the desired results in identifying vulnerable individuals.
