Children hooked on AI chatbots have markedly worse mental health

Schoolwork and emotional support

Developmental psychologist Anne Mahe, who was not involved in the study, told Oregon Public Broadcasting that children now turn to AI for emotional support almost as often as they do for help with homework. In her observation, hardly anyone was using AI just a few years ago.

For some students, AI is not only a study aid or a way to cheat, but also a conversation partner that helps them cope with anxiety. Arjun Sharda, 15, from Texas, said that before one of his first chemistry tests—on unit conversions and moles—he was afraid of failing and asked ChatGPT whether everything would be all right and whether his grade-point average would hold.

More than one in ten students, or 12%, use AI for friendship and conversation. A survey conducted last year found that nearly one in five high school students had tried to pursue a romantic relationship with AI, either themselves or through a friend.

The connection to mental health

JED Foundation researchers compared patterns of AI use with measures of well-being and mental health, as well as histories of suicidal thoughts and attempts. Kayla Anderson, a psychologist and the foundation’s vice president of research and evaluation, said that young people who turned to AI while emotionally distressed were more likely to report problems with their mental health and well-being, as well as suicidal thoughts and attempts.

Anderson noted that the available data are not enough to establish cause and effect. It remains unclear whether AI use worsens mental health problems, or whether teens already experiencing difficulties are more likely to seek help from chatbots.

Mahe believes both may be true: teens with problems may turn to AI for support, while interacting with a chatbot may make their condition worse instead of helping.

Adolescents need appropriate advice as they grow up, and people they can safely talk to about their feelings. A chatbot may seem like a trusted conversation partner, but its responses can sometimes be inappropriate or dangerous, and excessive agreement with the user can reinforce anxious thoughts. In some extreme cases, people have died by suicide after prolonged emotional conversations with a chatbot.

Dataist's take: The survey shows a link between emotional use of chatbots and problems, but does not explain what causes that link. Its findings therefore do not support the claim that AI itself worsens teenagers’ mental health: existing difficulties may also play a role.