A Word List With 49 Suicide Risk Factors, Used to Score Crisis Text

When someone reaches out for help in a mental health crisis, counselors have to make fast, imperfect judgments about who is in the most danger. A tool built by scientists at MIT's McGovern Institute for Brain Research works on the premise that the distressed person's own language carries clues worth reading systematically. Rather than training a system to guess at mood, it leans on a custom-built list of words and phrases tied to 49 suicide risk factors, searches text for those markers, and uses what it finds to estimate that person's risk. Daniel Low, a former graduate student in Satra Ghosh's group and now a research scientist at the Child Mind Institute, worked on the project with Ghosh and colleagues. In the Journal of Psychopathology and Clinical Science, the team reports that the tool accurately predicts suicide risk from text conversations, and the authors note that with more validation it could help with risk assessment in clinical settings and crisis-support situations. A dictionary of risk factors has an advantage over an opaque model in one respect: you can see exactly which phrase moved the score, which is the sort of thing a counselor might actually want to argue with.