- Thursday, April 25, 2024
The researchers’ preliminary findings suggest that a person’s social media friends and followers could indicate the likelihood of them developing depression
By: Kimberly Rodrigues
Researchers at the University of São Paulo are working on developing prediction models for anxiety and depression using Twitter and artificial intelligence (AI). The models aim to detect signs of these mental illnesses before clinical diagnosis.
The researchers’ preliminary findings suggest that a person’s social media friends and followers could indicate the likelihood of them developing depression.
The study involved constructing a database called SetembroBR, which included a corpus of 47 million publicly posted Portuguese texts and a network of connections between 3,900 Twitter users.
The research is published in the Language Resources and Evaluation journal.
During the Covid-19 pandemic, researchers manually collected timelines of tweets from about 19,000 users who had reportedly been diagnosed with or treated for mental health problems before the survey.
The study used two datasets, one for users with mental health diagnoses and another for control purposes, selected at random to distinguish between people with depression and the general population.
Ivandre Paraboni, a professor at USP and the last author of the study, explained that this was the first step in creating the anxiety and depression prediction model.
The study collected tweets from friends and followers of people with mental health problems, as they tend to follow certain accounts related to mental health.
Deep learning was used to create text classifiers and word embeddings based on bidirectional encoder representations from transformers (BERT), a machine learning algorithm employed for natural language processing (NLP) to detect the likelihood of a person developing depression based on their social media connections. The findings are still in progress.
Researchers found that BERT was the best model in predicting depression and anxiety by monitoring sequential data relationships such as words in a sentence.
People with depression tended to write about themselves and use topics such as death, crisis, and psychology.
Also, according to Paraboni, the signs of depression on social media may differ from those observed during a doctor’s visit.
For example, the study found that depressed users often used the first-person singular pronouns “I” and “me” and the heart emoji, which could be considered a symbol of affection and love. These observations could potentially be used as indicators of depression.
“This is widely felt to be a symbol of affection and love, but maybe psychologists haven’t yet characterised it as such,” Paraboni said.
The researchers are now working on expanding the database, improving their computational methods and enhancing the models to develop a screening tool for identifying individuals at risk of mental health issues.
The tool may also assist families and friends in detecting depression and anxiety in young people.
(With inputs from PTI)
