BUSINESS INSIGHTS : The latest news, analysis, and trends about protection and care

The MIT has developed an artificial intelligence tool designed to predict depressive tendencies through text or audio recordings. With an accuracy rate of 77%, this neuronal-network model represents a significant advance for AI in the complex and little-known field of human emotions. However, no specific applications have been announced for the near future.

A team of researchers from the famous Massachusetts Institute of Technology (MIT) has managed to develop a neuronal-network model that is able to detect signs of depression using text and audio recordings alone. A press release published on 29th August explains how artificial intelligence is gradually learning how to decipher human emotions, reports BFMTV.

 

77% accuracy

 

The artificial intelligence tool developed by the MIT does not require any context to make its diagnosis. It can detect a potential depressive state on the basis of just seven answers formulated in writing, or thirty for a voice recording. By detecting the patterns indicative of depression, such as specific words and intonations in the person’s speech or writing, it can predict whether the individual is depressed without the need for any further information or specific questions.

Tested on 142 people so far, the tool is around 77% accurate, claims the institute. It is a promising advance in the field of artificial intelligence, but one for which few concrete applications are planned in the near future.

 

Emotions, a little-explored field

 

Back in 2016, the MIT’s magazine – the MIT Technology Review – reported that American researchers were attempting to produce algorithms that would identify depressive individuals through their Instagram posts. Unfortunately, their results did not prove conclusive. The experiment just led to the observation that people suffering from depression are more likely to post darker photos than the average person, and photos whose main colours are grey and blue.

Prior to this, an Italian study published in October 2015 showed that, in bipolar subjects, extreme mood swings could be detected by smartphone use. Monitoring their location and their online activity could help identify their transitions from hyperactivity to depression, which are characteristic of this disorder.

 

Allianz Partners

Contact
Linked Topics
About Allianz Partners Business Insights