Using Machine Learning To Track Mental Health During COVID

Mental health has been a notable concern during the COVID pandemic that has placed such tremendous strain on people, whether due to concerns about the virus itself or the economic consequences of the lockdown measures designed to slow its spread.

New research from MIT and Harvard highlights how AI can be used to truly understand the toll the pandemic has had on our mental wellbeing.  They suggest that machine learning can be successfully used to monitor our language online, and therefore gain a handle on our state of anxiety.

The researchers analyzed over 800,000 posts on the Reddit website to understand how the changes in tone and content of language during the first wave of the pandemic may provide an indication of deterioration in mental health.

“We found that there were these natural clusters that emerged related to suicidality and loneliness, and the amount of posts in these clusters more than doubled during the pandemic as compared to the same months of the preceding year, which is a grave concern,” the researchers explain.

Early signs

The research also found that different types of mental illness had a different impact.  Collectively, the researchers believe their work could help not only those in the mental health field, but even moderators of online fora, such as Reddit.

“When the mental health needs of so many in our society are inadequately met, even at baseline, we wanted to bring attention to the ways that many people are suffering during this time, in order to amplify and inform the allocation of resources to support them,” they explain.

The researchers assessed content from 15 subreddit groups, each of which was devoted to some form of mental illness.  To provide a control, they also assessed groups unrelated to mental health, such as those concerning fitness, parenting, and personal finance.

The content was analyzed using natural language process to measure the frequency of words associated with topics such as death, substance abuse, anxiety, and isolation.  These words were then grouped together based upon the similarities in the language used.

Rising anxiety

The analysis revealed that conversations around COVID itself began, understandably, in March, the discussions around health anxiety began as early as January.  The frequency of discussions around mental health grew throughout the pandemic period, with even groups discussing things such as personal finance also heavily touching upon mental wellbeing due to the economic stresses caused by the pandemic.

The groups revealed not only that people were worried by the pandemic itself, but the restrictions put in place as a result of the pandemic also exacerbated problems such as eating disorders and ADHD, as people were deprived of important support systems.

The researchers then used a separate algorithm to group posts into clusters around topics such as substance abuse and loneliness.  They then tracked changes in these topics as the pandemic unfolded.  This analysis showed that discussions around suicide doubled from pre-pandemic levels.

While obviously it’s impossible to say that the pandemic is the sole cause of the changes seen, the research builds upon previous work from the team, which allows them to notice the clear changes over previous years.

It’s an approach that the team hope could prove useful in providing those involved in mental health with early warning regarding those groups most vulnerable to declines in their mental wellbeing, whether due to COVID or other stressors.

Facebooktwitterredditpinterestlinkedinmail