Can you tell the progress of depression with ChatGPT?

The prevalence of depression is a major social health problem, leading to crime due to depression, and there is a growing need to develop tools to predict who will develop depression. Previous studies have shown that analyzing a person’s word choice can predict worsening symptoms of major depressive disorder. In particular, research results have shown that AI can be of great help in these methods.

A study published in the Proceedings of the National Academy of Sciences (PNAS) found that ChatGPT can accurately predict the likelihood of worsening symptoms of major depressive disorder by analyzing the words a person chooses.

Researchers at Yale University had 467 participants fill out nine open-ended, neutral, short-answer questions and the Patient Health Questionnaire (PHQ-9), which assesses the severity of depression. Three weeks later, all participants completed the PHQ-9 questionnaire again.

The researchers used a tool called Linguistic Inquiry and Word Count (LIWC) to count the number of words in a particular category and determine how many words had a positive or negative emotional tone in participants’ written responses to short answer questions. . Additionally, human raters and ChatGPT versions 3.5 and 4.0 were asked to predict the progression of depression symptoms.

The study found that LIWC scores were related to the severity of depression at the time participants answered the questions, but did not predict severity of depression three weeks later.

On the other hand, emotional scores given by human raters helped predict future depression symptoms. “This shows that human raters are picking up on something that simply counting emotion words can’t,” explained Rob Rutledge, Ph.D., assistant professor of psychology in the Yale School of Arts and Sciences and lead author of the study.

Additionally, ChatGPT versions 3.5 and 4.0’s scores assessing the positive and negative tone of participants’ responses predicted future changes in depression severity very similarly to scores from human raters. “ChatGPT, which aims to mimic human conversational speech, takes word order and intra- and inter-phrase meaning into account in a way that standard language analysis tools such as LIWC do not,” the researchers said.

“Analysis of the language people use provides additional information that clinicians do not currently have, and our approach could be one way for clinicians to evaluate their patients,” said co-author Dr. Jutta Juman. He said.