AI that detects mental illness from postings on online bulletin boards is born



A research team at Dartmouth College in the United States has developed an AI that detects mental illness from posts on Reddit, the largest social bulletin board in the English-speaking world.

Emotion-based Modeling of Mental Disorders on Social Media Emotion-based Modeling of Mental Disorders on Social Media --2201.09451.pdf
(PDF file)

https://arxiv.org/pdf/2201.09451.pdf



AI Model Detects Mental Disorders Based on Web Posts | Dartmouth
https://home.dartmouth.edu/news/2022/03/ai-model-detects-mental-disorders-based-web-posts

According to the WHO, the proportion of people with mental illness affected by demographic transition has increased by 13% over the decade 2007-2017, with 20% of adolescents living with mental health problems. However, due to financial circumstances, social imprints, and low awareness of medical services, very few people seek help with mental health issues.

According to Xiaobo Guo, who developed an AI that detects mental illness from a post on an online bulletin board this time, if digital screening can detect people with potential mental illness, it will be possible to improve the treatment rate of mental health problems. It was said that. And SNS was considered to be suitable as a screening target in that it publishes its own data so that people can easily grasp their behavior. The reason this research chose Reddit from a wide variety of social networks is that it has nearly 500 million users and is discussing a wide range of topics.

The research team focused on the types of psychiatric disorders that have a clear effect on emotions, such as major depressive disorder, anxiety disorder, and bipolar disorder, and self-reported that they had any of these psychiatric disorders. We collect data on two types of users: users and users who do not have such mental illness. We have made it possible to analyze various basic emotions such as joy / anger / sadness / fear / no emotions that can be read from each post, and complex emotions that combine these basic emotions with AI.



According to the research team, there are 'patterns' in the emotional transition between posts, depending on the person's mental illness. The research team was able to detect whether a particular user has a mental illness by comparing the above-mentioned AI-read emotional transition patterns with the results of people with mental illness.

It is said that this research method is unique in that it analyzes 'emotions'. Most of the existing screening studies using AI focus on 'post content'. For example, the word 'new coronavirus infection' is associated with 'sadness' and 'anxiety' in AI that analyzes post content. If you learn that you are, the researchers who published the results of your research on the new coronavirus infection are determined to be sad or anxious. This method focuses on emotions, so it is free from such misdiagnosis.

The research team says that they will continue similar research to clarify emotional transition patterns.

in Science, Posted by darkhorse_log