To measure the impact of the Covid-19 pandemic on the mental health of the population, MIT researchers came up with the idea of using artificial intelligence to analyze 800,000 texts published on a series of Reddit discussion threads .
About fifteen of these threads dealt directly with mental health, focusing on subjects such as schizophrenia, depression or bipolar disorder. Eleven other sons were related to other topics, such as fitness, money, jokes or parenthood.
Artificial intelligence has been trained to detect words and phrases that bring out anxiety, fear of death, feelings of isolation, or illicit substance abuse.
Result: the researchers observed a very clear increase in these semantic changes between January and April 2020, which corresponds to the first wave of Covid-19. The focus group with the most negative terms was, unsurprisingly, the one on money and personal finance.
Among the mental health focus groups, the most negative were those dedicated to attention deficit and eating disorders, probably because these people no longer had access to the usual help and support structures.
More generally, messages referring to suicide have more than doubled over this period. They are particularly present in discussion groups relating to post-traumatic stress and personality disorders.
This study shows that it is possible to automatically assess the level of mental health in certain categories of the population. She also emphasizes the importance of a tool like Reddit in providing help and support to those in pain.
Researchers now want to create an alert tool that can detect mental health issues in online publications and, where appropriate, provide help and support.
Source : MIT