Scanning social media for suicide predictions

Posted Jan 2, 2018 by Tim Sandle
The Canadian government to set up a scheme whereby social media profiles and messages are searched and analyzed using an artificial intelligence tool, in order to predict suicides.
While it is impossible to guarantee that someone is or is not at risk of suicide, especially arising from a given mental health condition, research does suggest that where people with depression or certain suicidal thoughts can be caught early enough and measures taken by health agencies, then such actions can help to minimize the likelihood of a suicide occurring.
Working on the assumption that identifying tendencies early enough, the Canadian government is to invest in artificial intelligence technology so that social media activity can be screened with a view to early detection and the provision of support for particular communities.
The company selected for the analysis is Advanced Symbolics Inc. ASI’s artificial intelligence platform is named Polly and it has been operational since 2012. The primary application of Polly since the system's inception has been with market reteach.
According to CBC News, Polly will now be used to learn, and subsequently define "suicide-related behavior" on social media. The system will then "use that classifier to conduct market research on the general population of Canada." This is based on the success of a pilot that was running for the final three months of 2017.
The focus will be on trends rather than with pinpointing individuals (the latter practice crosses a privacy barrier that is not being considered as part of the current scheme). The trends can be used to inform social policy and to develop support services for certain regions, demographic groups and so on.
With trends, the tragic case of three teenage suicides on Cape Breton Island is highlighted as a reason why such a system is needed. It is hoped that future trend analysis will pick up such patterns and allow social services to respond proactively.