Scientists Develop Emotion Detection AI to Detect Signs of Mental Illness in Time

In fact, the use of artificial intelligence systems to detect mental illness is not new, and in April this year, the media revealed an analysis of user conversations on the social media platform Reddit to determine if there were any problems.

A team of computer scientists from Dartmouth College began training an artificial intelligence model to analyze social media text messages. The team chose the social media platform Reddit to train their AI model because Reddit has 500 million active users who regularly discuss a large number of topics on a network of Reddit branches.

The artificial intelligence system focused on finding emotional intent in network posts, rather than actual content information, and longitudinal statistics over time to track and analyze the mental health of the platform's users. In this study, staff focused on people's mood disorders -- major depression, anxiety and bipolar disorder, among others -- and these signs of illness were unique emotional patterns that could be tracked and analyzed.

The staff looked at data related to users who claimed to have a mental illness and compared it to data from users without any known mental illness, and they trained the AI model to flag the emotions expressed in users' posts and record the shifts in emotion between posts in a matrix plot. A post might be labeled by the AI system as "happy," "angry," "sad," "fearful ", "no emotion", or a combination of emotions, and the matrix shows how likely a user is to shift from any one state to another, e.g., from "angry" to A neutral state of "no emotion".

The team explained that different mood disorders have different patterns of mood shifts, and by creating an "emotional fingerprint" of the user and analyzing it against established mood disorder profiles, the model can detect signs of mental illness.

To validate the AI's results, staff had the AI system examine unused web posts during the training period, and the results showed that the AI accurately predicted which users may or may not be suffering from these disorders, and that the AI system would continue to improve and upgrade over time.