A recent study showed that artificial intelligence systems are now able to detect indicators of depression by analyzing texts published online, with an accuracy of up to 96%, in a step that may pave the way for the use of technology as an aid to monitoring mental health through digital platforms.
The research was conducted by a team from Georgia State University, and its results were published in the journal Corpus-Based Studies Across Humanities. The researchers used machine learning techniques to analyze thousands of posts on Reddit, a forum where users freely express their feelings and experiences.
“Language is not just a means of communication, but rather a mirror of inner feelings, and even when people do not declare that they are depressed, the way they express it can reveal their psychological pain,” said lead researcher Young-Min Kim.
The study relied on analyzing more than 40,000 posts from two different groups: the first includes users talking about mental health, and the second deals with general topics such as relationships and daily problems. The results showed that users of depression groups used first-person pronouns such as “I” and “myself” more frequently, along with expressions expressing loss of hope, which reflects an excessive focus on the self and a feeling of isolation.
The researchers also used topic modeling techniques based on large linguistic models (LLMs) to identify recurring patterns in texts. Topics related to family and education emerged, along with striking observations about social occasions such as birthdays and Christmas, as these occasions were associated with feelings of sadness or loneliness for some.
“Although the holidays represent moments of happiness for many, our data shows that for some, they can be painful times, increasing their feelings of loneliness or psychological stress,” Kim said.
Participating researcher Ute Romer-Baron confirmed that the project began as an academic mission before turning into a published scientific study, noting that the combination of applied linguistics and artificial intelligence tools allowed the analysis of natural language on an unprecedented scale.
Kim added that the goal of the study is not to diagnose patients, but rather to develop early warning systems that can detect the first signs of depression through digital activity, which supports the work of doctors and specialists in psychological care.
She concluded by saying: “Artificial intelligence cannot replace humans, but it may be an effective tool to identify those who need support before their condition worsens.”
The researchers plan to expand the scope of the study in the future to include anxiety disorders and post-traumatic stress disorder, comparing results across different languages and cultures.
