Can artificial intelligence like ChatGPT detect early signs of depression?

Artificial intelligence algorithms such as ChatGPT have the potential to revolutionize the world of medicine, particularly in the field of mental health. There are already studies aimed at establishing whether this AI can detect early signs of disorders such as depression.

Early detection of such disorders is essential to ensure proper treatment and patient well-being. However, in many cases this is the least understood and difficult to detect stage of the disease. That's why a group of researchers at the University of Victoria in Australia have taken it upon themselves to integrate chatbots like ChatGPT into the psychiatric process.

To this end, a custom ChatGPT-style chatbot was developed, calling it DEPRA ("Depression"). With this tool, the researchers are looking to establish whether this class of AI can serve as a support when it comes to identifying and helping patients at the earliest stages of the disease.

DEPRA can help members explore their own mental health as a guideline so health professionals can help patients with depression.

The researchers used the Google Dialogflow platform to create the DEPRA chatbot. With its interface, they were able to create a kind of ChatGPT to diagnose depression in a selected sample of the population. In addition, artificial intelligence was also able to filter the results by their levels of severity based on the Hamilton Depression Scale.

DEPRA is integrated into Facebook* Messenger, the platform members used to communicate with DEPRA. After that, the chatbot started a user survey, which took about 30 minutes to complete. The study noted that interaction with artificial intelligence brought “high levels of satisfaction” to participants.

The chatbot's artificial intelligence takes into account "mood, guilt, suicidality, insomnia, arousal or lag, anxiety, weight changes, and bodily symptoms to diagnose levels of depression," they describe in the study. To test this, a group of 50 Australians was recruited.

They indicated that the questions were easy to understand and easy to answer, that it took them less time than attending a face-to-face session with a psychiatrist, and that they preferred the option of texting via social media platforms rather than talking to a psychiatrist on consultation session.

For researchers, using artificial intelligence to diagnose depression early has an added benefit. By communicating directly with the computer, patients may be less embarrassed to speak, expressing their thoughts openly. The 24/7 availability of the tool also adds additional flexibility not found in current methods of working with mental health professionals.

Promising results from DEPRA

According to the results obtained with the help of artificial intelligence, 30% of respondents did not have symptoms of depression. For this reason, the DEPRA chatbot has decided not to recommend a session with psychiatrists or other medical professionals. However, 20% of the participants reported very severe symptoms of early depression, while 22% showed moderate symptoms.

The research team plans to improve their DEPRA chatbot in the future. While AI currently relies on manual calculation based on the reaction of the user it assigns a score to, the next version will use machine learning and sentiment analysis algorithms to automatically calculate depression levels closer to technologies like ChatGPT. In this way, participants can get results in the shortest possible time and get the help they need.

Of course, there is still a lot of work to be done, since AI like ChatGPT is relatively new and the knowledge about it is not that deep yet.

* - belongs to the company Meta, banned in Russia.