Researchers at MIT's McGovern Institute have developed a language-processing tool to evaluate suicide risk from text conversations. Published in the Journal of Psychopathology and Clinical Science, the new study utilized de-identified text conversations from approximately 16,000 crisis counseling sessions provided by Crisis Text Line to help professionals better assess mental health emergencies.
How does the new suicide risk tool work?
The language-processing tool relies on a custom-built lexicon linking words and phrases to 49 specific suicide risk factors. A machine learning model then scans text conversations to identify key predictors, such as active suicidal ideation, substance use, and mentions of lethal means, giving human counselors explainable risk assessments.
Why is this machine learning model important?
Unlike heavy computational systems, this lightweight model can run efficiently on a personal computer while maintaining data privacy. By highlighting exact words of concern, it provides human counselors with interpretable insights during critical moments of a mental health crisis.
