Sylvie Bernaerts, PhD, from the Center of Expertise Care and Well-being at Thomas More University of Applied Sciences in Belgium, shared insights from her recent research on the use of artificial intelligence chatbots in mental health. The study investigates how individuals utilize these chatbots for personal and interpersonal issues, as well as the implications for clinicians and privacy concerns.
Read More
Bernaerts explained that the project was prompted by a lack of knowledge regarding the extent to which people in Belgium use AI for mental health support. The research sought to fill gaps in understanding user behaviors and motivations to better inform mental health practitioners.
The findings suggest that users often discuss personal matters with AI chatbots, indicating a demand for support that does not always align with traditional therapeutic needs. Specifically, younger users, such as students, often sought quick advice about conflicts or emotional reassurance rather than professional care. This raises concerns about people using chatbots as substitutes for diagnosis or therapy, a practice that can be risky even if it appears to be a minority behavior.
The study revealed that many participants preferred accessible AI tools, notably ChatGPT, due to their availability. Bernaerts advises clinicians to engage with clients about their chatbot use to learn what assistance they have found helpful and to address any misinformation they may have received.
Additionally, a considerable portion of those surveyed were already receiving professional mental health care. Bernaerts envisions a complementary role for specialized AI chatbots in mental health care. For example, they could assist with intake processes, provide exercises during treatment, and aid in relapse prevention post-treatment. However, she emphasized that currently available chatbots aren't adequately validated for these roles.
The analysis also explored user engagement and the digital working alliance with AI chatbots. Results indicated that individuals who interacted frequently with chatbots reported greater trust and rapport, though the study could not determine causality in this relationship.
From both clinical and ethical perspectives, concerns were raised about the potential for AI chatbots to serve as substitutes for human interaction, especially given the ongoing challenges of access to care. Bernaerts highlighted that while some safeguards exist for general-purpose AI chatbots, more robust measures are needed, particularly in mental health contexts. These chatbots should avoid excessive compliance and instead offer structured interactions based on established therapeutic models.
Looking forward, Bernaerts emphasized the importance of addressing key research questions regarding the integration of AI chatbot support into standard care. Understanding clinicians' perspectives on operational opportunities and barriers is essential for successful implementation of this technology.