ChatGPT's new teen mode may help kids learn but weaken their own thinking

OpenAI has automatically enrolled every ChatGPT user under 18 in a new Teen Mode, launched in August. It adds safety limits and a study tool that makes students do much of the thinking. In a Guardian opinion column, psychoanalyst Robin Kirman argues that even the helpful part carries a cost: teens may not get to build trust in their own minds.

What Teen Mode does

The launch came a year after a lawsuit alleged that ChatGPT contributed to the suicide of a 16-year-old, Adam Raine. According to the column, Teen Mode includes these rules:

• Guardrails against destructive content
• Parent notifications for self-harm concerns
• A ban on posing as a friend, claiming feelings, or suggesting it is more than a tool

The last rule responds to a case involving Character.AI, in which a 14-year-old formed an intense attachment to a chatbot.

The study tool

Study Mode is built into Teen Mode. It refuses to hand over answers and asks the student to work them out. OpenAI's Leah Belsky describes it as a shift from "a straightforward answer engine into a cognitive tutor." The column notes research suggesting that AI that simply does kids' work lifts grades briefly, then the gains collapse once the tool is gone. Kirman calls Study Mode's approach genuinely new, and a real help for students with no tutor.

The worry

Kirman's concern is that the tool is too good. It never gets tired, never goes home and never runs out of patience. She points to the British pediatrician and psychoanalyst Donald Winnicott, who argued that children need a "good-enough" parent, not a perfect one. A limited parent forces a child to find her own resources.

Teenagers, she writes, need to pull away from the minds they depend on and find what they actually believe. Part of that is simply feeling sure they are right. A teen whose thinking is always guided, she argues, never becomes her own authority.

She expects OpenAI to answer that Teen Mode is a scaffold, not a crutch. She doubts teens will ever have a reason to let go of it. She also says what grows with support is not the same as what grows without it.

What is not known

This is one clinician's argument, not a test of Teen Mode. The material gives no results showing that Study Mode weakens teenagers' thinking, and no data on how teens use it.

What to do

Parents can ask their teen how they use ChatGPT, and whether they try a problem alone first. Setting aside some homework and some ideas to work out without any assistant keeps room for the struggle Kirman says teens need.