An Investigation into the early detection of psychosis and depression in individuals via Social Media (Twitter)

In this paper, we describe a pilot study undertaken to analyse Tweets that identify possible online signs of distress. 20,800 tweets were collected between July and November 2013 that contained potential indicators of depression and psychosis. In order to identify tweets that contain signs of distress, a subset of 2,500 tweets was judged by the author as to whether they contain ‘worrying’ phrases. In order to establish levels of inter rater reliability, 100 of these tweets were then assessed in the same way by 3 independent judges with a mental health background. The results suggest that Twitter users feel comfortable expressing low self esteem issues as well as mental health difficulties to online friends, family and other Twitter users