MIT researchers test lightweight suicide risk tool on crisis texts

MIT researchers reported that a lightweight model estimated suicide risk in previously unseen crisis text conversations by looking for language tied to 49 risk factors. They developed and tested it using de-identified texts from approximately 16,000 conversations with Crisis Text Line counselors. The model can run on a personal computer and show which terms contributed to its assessment. Mentions of lethal means and substance use were more common in the highest-risk group than mentions of depressed mood or fatigue. The researchers say the tool needs further validation before clinical use; its word lists can miss context or unfamiliar expressions.