Developing Accurate Internet Suicide Prediction Algorithms

Can artificial intelligence (“AI”) help Internet sites detect suicidal people? Apparently, some psychologists think so. Today a number of academic researchers and high tech entrepreneurs seek to create algorithms to enable computers to pinpoint and identify posts by people who may find themselves considering suicide using AI tools.


Dr. Jessica Ribeiro, a professor of Psychology at Florida State University, recently led a study which concluded artificially intelligent computers can predict suicide attempts with between 80% and 90% accuracy as early as two years in advance. One week prior to a suicide attempt, the accuracy rate climbs to an astonishing 92%. Another study from researchers at the same institution revealed that during the past five decades, human beings had not succeeded in predicting suicide accurately. Clinicians typically focused on three high risk factors: stress, depression and substance abuse.


Facebook has reportedly been working along the same lines. A news story carried by the BBC indicated the large social media company has already started using an algorithm coupled with artificial intelligence to screen postings. When the tech company identifies account holders at high risk of committing suicide, it seeks to contact them to suggest alternatives. The BBC news story claimed Facebook has begun testing this new technology in the United States.


Representatives from organizations committed to suicide prevention assisted Facebook in refining its pattern recognition tools to identify potentially at-risk posters. The new app reportedly offers several support options to possibly suicidal posters. Some critics have expressed concern the new suicide identification app, which cuts off live video streams, may actually remove a useful tool for obtaining timely intervention. In the past, upset Facebook users reportedly contacted the police about posts threatening suicide and prevented a number of deaths. Research into this complex area remains ongoing.