ChatGPT’s teen mode may weaken adolescents’ sense of independence

What Teen Mode changes

Technology companies have spent years disputing links between their products and the harms alleged in similar lawsuits. The suits claim that companies not only failed to protect children but deliberately built products capable of exploiting their vulnerabilities. In Kirman’s view, Teen Mode is an indirect acknowledgment that AI behavior itself can cause harm, not just the content of its responses.

Kirman is a psychoanalyst who works with young people. She observes how social media and AI affect not only adolescents’ behavior and moods, but also the development of their identities and characters.

Teen Mode is designed to limit obvious risks. It includes safeguards against harmful content and notifies parents if the chatbot detects concerns about self-harm. It is also not supposed to act like a friend, claim to have feelings of its own, or present itself as anything more than a tool. The last restriction came against the backdrop of a wrongful-death case involving a teenager who had become deeply attached to the Character.AI chatbot.

Study Mode and the limits of assistance

Study Mode, built into Teen Mode, is meant to help adolescents learn rather than simply give them ready-made answers. Kirman says recent research has confirmed that when AI does students’ assignments for them, their grades improve temporarily, but the effect disappears once the tool is taken away. Ultimately, those students are worse off than students who did not use AI.

OpenAI vice president of education Leah Belsky describes Study Mode as a shift from a simple answer-delivery mechanism to a cognitive mentor. The mode follows a Socratic approach: instead of giving the solution, it asks the student to do a significant part of the mental work themselves.

For a teenager without access to a tutor or a teacher who has time for their questions, such a conversation partner could make substantial intellectual support more accessible. It can answer patiently and cover more knowledge than any one person. Unlike a tutor who goes home, or parents who get tired and have other things to do, AI can keep the conversation going indefinitely.

Kirman sees a problem in this constant availability: Teen Mode becomes “too good a parent.” British pediatrician and psychoanalyst Donald Winnicott, who spent decades studying the formation of a child’s self, believed that children need an environment that is not perfect, but good enough.

A “good enough” parent is present and gives a child support, but remains limited: they can become distracted, tired, or make mistakes. The child then has to rely on their own abilities and discover that they can manage. A perfect parent would spare them that struggle—and the sense of self that it helps build.

The argument applies to adolescence, too: this is a time when it matters to separate from other people’s views and seek out one’s own. Psychoanalyst Peter Blos described growing up as moving away from the authority of adults whose opinions adolescents once relied on, and trying to understand what truly belongs to them. Doubting parents, teachers, and accepted beliefs can be part of that process.

An adolescent does not necessarily have to be right. What matters is the confidence that they are right: it gives them the courage to find out what they actually believe. If a conversation partner is always there in an adolescent’s intellectual life to steer them and supply an answer, it becomes harder for them to become an authority in their own right.

When support gets in the way of independence

One could argue that Teen Mode is a temporary support, not a lasting dependency: an adolescent will learn to reason through conversations with AI, absorb that way of thinking, and gradually use the tool less. But Kirman says it is unclear whether adolescents will have any incentive to give up the assistance, especially as AI becomes more widespread.

She also questions whether development with this kind of support can be considered equivalent to development without it. Kirman says many technology executives send their own children to schools where students can complete assignments without AI assistance. Teenagers without those options will get Teen Mode—or, worse, AI without Socratic constraints.

Kirman believes the stakes include not just cognitive skills, but also a person’s ability to understand their own thoughts reliably.

Other areas have already shown how constant connection to technology affects a developing personality. Jonathan Haidt synthesized research on smartphone use and showed that the sharp rise in anxiety, depression, and suicidal thoughts among adolescents coincided with technology entering their free time without adult supervision. In his account, technology displaces the private space where identity takes shape.

Sherry Turkle, who has studied adolescents and technology for decades, found that many say they do not know what they think until they post it and see how others respond.

In her practice, Kirman meets young people who struggle to understand what they feel and want. She sees this instability as part of the youth mental health crisis. A person without an internal compass becomes anxious, seeks confirmation of their feelings, doubts them, and easily loses their balance.

Kirman believes Teen Mode does not address this deeper problem. It asks adolescents to think while continuing to guide them and reinforce their dependence on external help. A stable sense of self requires moments of real struggle, when a person works through difficulty and discovers they can withstand it. One such experience is a blank page, with a person alone with their doubts, anxiety, discoveries, and joy.

The German philosopher Arthur Schopenhauer wrote in the 1850s that thinking for oneself is a little uncomfortable, which is why few people do it. Kirman fears that a new generation will face this necessity less and less often. The loss is difficult to measure and therefore easy to miss, but in her view, it is through the discomfort of thinking independently that people discover they have minds of their own.

Robin Kirman is a psychoanalyst and writer. The closing note to the original publication also lists suicide-crisis support services in the United States, the United Kingdom, Ireland, and Australia, along with the international directory Befrienders.

Dataist’s take: The debate is not only about whether AI can explain material; it is about how much room adolescents have to find answers on their own. Study Mode offers more guided help than a chatbot that simply gives answers, but Kirman argues that constant guidance can hinder the development of independence. The source provides no data that would allow this long-term effect to be measured.