Researchers from the Massachusetts Institute of technology (MIT) recently developed a neural network which with a relatively high degree of accuracy predicts the probability that a person has cognitive impairment.
This development can also be called a depression detector. A team of researchers at MIT introduced the project this week at the conference Interspeech 2018 in India.
According to a press release, they developed a context-free method by which an AI can learn is written by a man text and assign a rating indicating the level of depression.
“The therapist uses a combination of proven questions and direct observations to diagnose mental health conditions such as depression. Artificial intelligence could do something like that without having to ask conditional questions and to conduct direct observation. It doesn’t need context – it’s enough just to get the data,” said the team from MIT.
To check AI researchers conducted an experiment, which was attended by 142 persons screened for depression. They answered a number of questions managed by the virtual agent. AI had preliminary answers to the questions.
The answers of the participants of the study were recorded in the form of text and audio. In the text version of the AI was able to predict depression after seven issues. But, interestingly, in the audio version of the intelligence it took about 30 sequences to determine. It averaged precision, according to the researchers, was amazing and made up 77%.
There is no reason to believe that the development of the MIT will be used by all who directly interact with patients. Therapists tend to believe that they are trained to diagnose patients better than the machine algorithm.
The use of this technology outside the laboratory can cause panic among the people because of the level of depression defined by AI, will depend on work, school, social status, and more.