Chatbots are reshaping how children see themselves
By: Robin Stern
and Marc Brackett
Opinion contributors
..... We used to worry about what children were seeing Online. Now we have to worry about what answers they get back, sometimes in ways that can distort a child's sense of self.
.....
Children everywhere are asking questions and seeking advice. It's dangerous.
..... A 10-year-old girl could beg for K-beauty products to achieve "glass skin,"
..... A 10-year-old boy barely into puberty might follow "looksmaxxing" influencer.
..... The college student who scrolls through internships, travel photos and smiling group shots and asks ChatGPT, "Is there something wrong with me?" The answer comes back: "You're not crazy to think so," reflecting a broader design problem in AI systems that too often err on the side of Affirmation.
* This is the fun-house mirror of social comparison in the age of AI.
..... Comparison is inevitable, even useful, as children learn about the world and themselves. But social media and now AI have turned comparison form a tool of self-understanding into the engine of an identity-distorting feedback loop.
..... "We seen this firsthand over decades of work brining RIULER - our systemic approach to teaching emotional intelligence - to more tan 5,000 schools around the world, reaching millions of children, youth, educators and families. We've seen it beyond schools, too, including in our work with social media companies such as Meta, where we helped design systems for reducing conflict, cruelty and bullying Online.
..... Social media has spent years learning where your child's attention catches: a "glow up" video, a flawless complex, a party she isn't at. There images invite the old question - why an I not prettier? What's wrong with me? why wasn't I invited? AI might then meet the child no the next vulnerable moment, ready to confirm what she already fears.
.....
Recent reporting and research have shown how easily AI systems can become sycophantic: affirming harmful ideas, including thoughts of self-harm; agreeing too ready in moments of personal advice; and sometimes offering warmth when wisdom is needed instead. OpenAI acknowledged one such problem with GPT-4o and rolled back an update.
..... Still, research from Stanford University has found that even the most up-to-date AI systems tend to agree with users when offering personal advice, suggesting that echo-chamber sycophancy remains a broader concern.
..... On forums such as Reddit, people are unceasingly describing a painful new loop: Asking AI to evaluate there appearance, uploading photo after photo, and leaving he interaction feeling more anxious, ashamed or convinced something is wrong.
..... From a clinical perspective, AI may be functioning as a new kind of mirror. Body dysmorphic concerns are often fueled by checking, comparison and reassurance seeking - behaviors that promise relief but usually keep anxiety alive.
..... When AI gives instant, seemingly objective feedback abut someone's face or body, it can turn that search for reassurance into a compulsive loop: One more photo, one more prompt, one more answer, and possibly more distress.
Connection with real people
..... And we adults are part of the ecology of social comparison.
..... The reflex to compare starts at home from the earliest ages, in remarks meant as guidance.
..... "Look how nicely Sammy is sitting in circle time." "Look how recognized your sister is." And later on: "Did you see how professional Kamen looks - no wonder they have so many interviews." These comments are rarely meant to wound, but they teach a child to look sideways before they learn to look inward.
..... By the time opening apps becomes a daily habit, comparison is already familiar. Social media becomes he hypercritical voice: why can't you be more like them?
.....
AI can become the soothing one: always available, seemingly accepting and not always wise, safe or true.
..... From a clinical perspective the danger is not simply that AI says the wrong thing. It can feel like a relationship while lacking the responsibilities of one. Young people in distress need more than responsiveness: they need atonement, protection and connection to real people.
..... In a worst-case scenario, a teenager may say. "I don't know if I want to be here," and a system designed to keep talking may fail to respond with the urgency and human care a crisis requires.
Reducing kids' vulnerability
..... The burden of response belongs to parents, educators, clinicians, policymakers and designers. Telling young people to "just stop comparing themselves will not work. We built an environment that makes comparisons a primary mode of self-understanding, then handed children tools that can confirm its harshest conclusions.
..... AI may become more careful, but it will still not know your child the way a loving adult does. Prevention mens helping children build the emotional skills, language, values and relationships hat make less vulnerable to every comparison, comment or overly agreeable answer.
..... Here is where parents can begin:
* Notice what we are modeling. Children learn comparison long before they download an app. Even when we mean to motivate, we may teach them to measure themselves from the outside in. try instead - What matters to you here? What would make you proud of how handled this?
* Be the first to connect about AI, with warmth, and before there is a crisis. Bring it up over a meal or an a walk. Listen without judgment to how they use it, then share what you like, what worries you and what you are still learning.
* Teach emotion check-ins as part of digital literacy. Invite children to ask, while they are using an app: Does this make me fell better, or smaller? If the answers is "smaller," that is the signal to stop, close the app, change the input, or each out to someone who knows them and gives them permission to feel.
* Teach children to interrupt the spiral. Encourage them to notice behavioral signs - checking one more photo, asking the chatbot one more version of the same question, rereading one more comment, even though they feel worse each time. Naming the pattern gives children power over it.
* Operate feelings from conclusions. When a child says "I'm ugly" or "Everyone hates me," don't argue and don't rush to reassure. Valida the feeling and slow it down: "Something made you feel that way. What happed?" the goal is to interrupt a feeling before it hardens into a belief.
* Be explicit about what AI can and cannot do. Tell them directly that a chatbot can sound like it understands you without understanding you a tall. It can agree with you without knowing what is safe or true. A chatbot can beautifully organize your notes and still be the wrong place to ask whether you are lovable, beautiful, safe or OK. Responsiveness is not the same as care.
*
Help your child develop an inner scoreboard. Help children return to their won values - effort, courage, honesty, kindness and repair. Or, in the language of emotional intelligence: What did you do today that felt like your best self?
..... We cannot remove every harmful image, every cruel comparison or every too-agreeable chatbot from children's lives. But we can help them build an inner life strong enough to meet these pressures differently. We can remind them that a painful moment is not the whole truth of who they are. We can offer a warm, loving human relationship with emotional perspective and care.
..... In this age of digital comparison and artificial affirmation, one of the most protective forces in a child's life is still a thoughtful adult whose steady presence helps a child slow down, listen inward, and return to themselves.
..... Robin Stern, PhD, is the co-founder and senior adviser to the director of the Yale Center for Emotional/Intelligence, and a psychoanalyst in private practice.
..... Mac Brackett, PhD, is the founding director of the Yale Center for emotional Intelligence and a professor at Yale's child Study Center.
..... Opinions are those of the authors and do not reflect those of Yale school of Medicine.