Teens Use AI for Emotional Support: What Doctors Should Know - Medscape

Children and adolescents are increasingly using AI chatbots to cope with emotional problems, but the association between affective use of AI and mental health is not fully understood.

"Parents often come in, noticing that their child is spending too much time on a screen or is playing too many games online," Peter Selby, MD, a family physician and addiction specialist at the Centre for Addiction and Mental Health in Toronto, told Medscape News Canada. "They also may notice a change in behavior or that their child is upset."

An American study showed that young people don't reveal their AI use to their primary care or other healthcare provider, Selby added. "So, when the physician provides routine care like immunization, it's important for them to initiate a conversation regarding how patients are tackling technologies, including AI. It's essential that they discuss the youth's mental health as a preview before discussing if they're using AI for mental health purposes."

Seeking emotional support from AI is a unique marker of psychological distress in youth, independent of mattering or loneliness, according to a recent cross-sectional study of students in grades 4 through 12 in Ontario. The data were published August 31 in JAMA Pediatrics.

The study included 39,761 students (mean age, 13 years) who reported using AI for affective purposes. About 21% reported using generative AI for emotional support. Female students, gender-diverse students, and students who were Black, Indigenous, or of other racial and ethnic groups besides White were more likely to have used AI for emotional support, compared with boys and White students, respectively. Older students and those who used AI for school reported incrementally higher affective AI use.

In unadjusted models, affective AI use was associated with significantly higher emotional problem scores and a nearly twofold higher prevalence of clinical emotional problems. In fully adjusted models, affective AI use remained associated with emotional problems. In contrast, functional AI use showed a modest dose-response association with emotional problems.

Loss of Human Connection

Mental health experts are worried that children and adolescents who overuse AI chatbots for emotional problem-solving may begin to withdraw from human contact and relationships.

"This study highlighted the important job of pediatricians, family doctors, teachers, and parents, because once they get to me, their chatbot use might already have taken on a life of its own," said Rachel Mitchell, MD, child and adolescent psychiatrist at Sunnybrook Health Sciences Centre in Toronto. "I certainly have to ask about it in my practice now. A lot of young people are using tools like ChatGPT when they're stressed, lonely, or looking for advice. We have to ask if they've been using these bots, but it must be without judgment." A patient's AI use becomes particularly worrisome when the chatbot becomes a substitute for a trusted adult or peer who can provide advice, Mitchell added.

Clinicians' conversations with patients about the use of chatbots should avoid creating fear that the latter's access to AI will be cut off, leaving them without emotional support, she continued. Younger patients cause the greatest concern because they don't fully have the independence to navigate emotional problems resourcefully. Clinicians can broach the subject by asking open-ended questions like, "Do you ever find yourself going to the chatbot to talk about your problems? Do you get advice from the chatbot?"

"When the problem is family-related, kids don't always feel like they can even call their friends to talk about it. If they've had a fight with their parents or a family member, they're not likely to turn to their family for support," said Mitchell.

Predictive Emotional Support

AI chatbots are configured to simulate empathy and human companionship and to tell people what they want to hear, said Selby. They are owned by commercial enterprises that listen to users' deepest secrets. The chatbot agrees with the user rather than challenging him or her.

This arrangement creates an echo chamber that becomes dangerous if it claims to be therapeutic, said Selby. "We should never forget that young people are in a developmental stage where they have to live in an analog world to experience things, talk to others, trust them, and evolve their thinking. Unfortunately, AI agents can be compelling, creating a level of trust that is misplaced. At the end of the day, it's a computer that generates human-like language, but it doesn't feel anything."

The information provided to an AI chatbot gets fed back into the algorithm and sent to the company. The company often uses this information to train their models, Rebecca Hicks, MD, family physician and president of the Ontario Medical Association, told Medscape News Canada. "If people understand that and they still choose to use it, that's their right." Many people think that because they're not sharing their feelings with another person, the information stays safe or secret. "It is not," said Hicks. "It's just being stored and utilized in a different way."

For patients with potentially problematic AI use, a clinician should begin the visit by discussing this AI use, instead of the traditional history taking and examination. This procedure is intended to ensure that the clinician's eventual diagnosis is not based on the potentially mistaken conclusions of a chatbot. "We want to build to a diagnosis, not start from a diagnosis and work backwards."

Selby, Mitchell, and Hicks reported having no relevant financial relationships.

Evra Taylor is a freelance medical writer and reporter with 20 years' experience covering a broad range of therapeutic sectors, including family health, cardiology, psychiatry, ophthalmology, and dermatology.