An artificial intelligence model created by Dartmouth researchers can gain insight into people’s mental state through their social media posts.
Algorithms have shown in recent years that they can be used both to help people and for control and extreme surveillance.
good examples of it are the algorithms capable of diagnosing all kinds of diseasesfrom depression to cancer. On the dark side, a Microsoft AI served an Argentine province to estimate which girls could get pregnant.
A pioneering AI model is capable of detecting mental disorders using conversations on Reddit. It has been created by Dartmouth researchers and instead of relying on text content as other projects do, his novel approach is based on emotions.
In the research paper, presented at the 20th International Conference on Web Intelligence and Intelligent Agent Technology, its leaders argue that it works better.
“Social media offers an easy way to access people’s behaviors,” says Xiaobo Guo, co-author of the paper.
The authors they selected Reddit because this news forum format platform has nearly 500 million users to its credit active, who use their space to discuss and share a wide range of topics, publicly available. Data was collected going back to 2011.
Once the information is obtained, the authors focused on emotional disorders (major depressive, anxiety and bipolar disorders), named after them because they present different emotional patterns.
After analyzing data from users who reported having a disorder and also from those without a diagnosed disorder, they trained the artificial intelligence to be able to recognize the emotions expressed in the posts and map the transitions between contents, labeled as “joy”, “anger”, “fear”, “sadness”, “no emotion”, or combined.
Emotional fingerprint for the user
As explained in the official statement, their algorithm was able to create an emotional fingerprint for the user and compare it to the distinctive pattern that characterizes each disorderso that the model can detect them.
When validating the results, they tested it on posts that were not used during training, confirming the accuracy of the AI-based model.
Other detection tools based on scrutinizing text content can be misleading. “If a model learns to correlate COVID with sadness or anxiety, it will naturally assume that a scientist who studies and publishes on the subject suffers from the disorder,” says Vosoughi, another co-author.
Those responsible hope that their study can provide new paths towards the prevention of mental disorders. According to WHO data, mental health problems like anxiety have skyrocketed during the pandemic.
In fact, in its latest report, the World Economic Forum in Davos places the problem as the sixth global risk for 2022. The Davos survey, in which nearly 1,000 experts, world leaders and academics participated, reveals an “omnipresence” of mental health ailments or disorders worldwide.
This article was published in Business Insider Spain by Andrea Núñez-Torrón Stock.
George is Digismak’s reported cum editor with 13 years of experience in Journalism