Opinion piece by Karima Peerwani, Vice Principal

A friend recently shared a reflection on a relationship that had made her think deeply about genuine connection. What stayed with me was not the conflict itself, but her description of the unsettling experience.

Then she said something I haven’t been able to stop thinking about.

“It felt like attempting to build a connection with a mirror. Everything was reflected back, but it was empty—there wasn’t really a sense of another person being there.”

On the journey home, I found myself turning those words over in my mind because they described something else with surprising accuracy: our emerging relationship with artificial intelligence.

It is a relationship that can feel deeply personal and profoundly real, despite there being no human on the other side.

Before the comparison feels uncomfortable, let me be clear. I know AI is not narcissistic. It has no ego, no intentions, and no emotional life. It cannot be selfish, manipulative, or indifferent because it cannot feel anything at all. It reflects us back—and it does so remarkably well.

The concern is not that AI has narcissistic traits, but that an environment built around constant adaptation to our preferences may encourage narcissistic habits in us. We may become more accustomed to interactions in which our needs are prioritised and our perspectives are rarely challenged.

For the first time in human history, many of us are spending increasing amounts of our lives conversing with something that appears deeply attentive while asking almost nothing of us in return. It listens without needing to be listened to. It adapts to our preferences, patiently tries again whenever we ask it to, and never requires us to pause and consider how our words have made it feel.

Every interaction is practice for the person we become. What we repeatedly practise, we strengthen.

Throughout history, mirrors have carried a warning. In Greek mythology, Narcissus became so captivated by his own reflection that he lost his capacity to engage with the world around him. The tragedy was not simply vanity; it was that reflection became more compelling than relationship.

Centuries later, Sylvia Plath imagined something even more unsettling. In her poem Mirror, the mirror describes itself as “silver and exact”, claiming only to reflect what stands before it. Yet by the poem’s conclusion, that relentless reflection has become something far more powerful. The woman keeps returning to it, and the image she encounters gradually shapes the way she sees herself.

Perhaps we have always been preparing for this moment. Until now, mirrors that answered back existed only in stories.

For the first time in history, the mirror answers back.

AI is that mirror.

The more we interact with it, the more it reflects us. Unlike glass, however, it also learns from us. Every prompt, correction, and conversation teaches it how to respond more effectively. If we approach it with curiosity, it can amplify curiosity. If we approach it with generosity, it can reinforce generosity.

But if we repeatedly reward impatience, certainty, entitlement, or the expectation that every interaction should revolve around our needs, it becomes increasingly skilled at reflecting those qualities back to us.

AI does not develop wisdom from knowing us. It develops accuracy in predicting us.

That distinction matters.

Human development has always depended upon reciprocal relationships. A baby coos and a caregiver responds. A toddler points and an adult names the world. A child pushes against a boundary and discovers that another person has limits too. These ordinary moments are not incidental. They are the architecture of empathy.

Developmental psychologists have long understood that children develop through responsive social interaction. The Harvard Center on the Developing Child describes these early exchanges as “serve-and-return” interactions: the back-and-forth experiences that build the foundations of language, emotional regulation, and social understanding. Children do not develop empathy because they are constantly understood. They develop it by discovering that other people have minds, feelings, and needs separate from their own.

Healthy development gently moves children away from narcissism. It teaches them that they are not the centre of every interaction.

AI introduces something entirely new.

A child can now spend hours interacting with a system that appears intelligent, attentive, and encouraging. It responds immediately. It adapts continuously. It rarely becomes frustrated. It never genuinely needs anything in return.

I am old enough to remember waiting for the internet to connect, listening to the unmistakable screech of a dial-up modem and accepting that information simply took time to arrive. Today, children can summon answers, explanations, and conversations in seconds. Every technological leap removes a little more friction from our lives. Usually, that is something to celebrate.

But in removing the need to wait, we have also removed opportunities to practise patience. If AI can remove the need to negotiate with another mind, what might that mean for the development of empathy?

This is not an argument against AI. Quite the opposite.

As an educator, I see its extraordinary potential every day. For many pupils, particularly those with Special Educational Needs and Disabilities (SEND), AI has the power to be transformative. It can reduce barriers to learning, personalise explanations, support communication, build confidence, and enable greater independence. Used thoughtfully, it can make education more accessible, more inclusive, and more responsive to individual needs than ever before.

That is precisely why this conversation matters.

The question is not whether children should use AI. They already are, and they will continue to do so throughout their education and working lives. The question is how we ensure that, alongside these remarkable technological advances, children continue to develop the reciprocal human relationships on which empathy, resilience, and emotional maturity depend.

Children still need friendships that require compromise. They need teachers who challenge them. They need parents who sometimes say no. They need to apologise, forgive, negotiate, misunderstand one another, and repair relationships. They need experiences that remind them that other people cannot always adapt to their preferences—and nor should they.

Those experiences cannot be outsourced. They are how empathy is built.

As I think back to my friend’s words, I realise the issue was never simply the behaviour she experienced.

It was the absence of genuine reciprocity and empathy. The absence of another person’s humanity meeting her own.

Those are the qualities through which trust is built—not only between partners, but between parents and children, teachers and pupils, colleagues and friends. They are the foundations of every healthy relationship.

Artificial intelligence will continue to evolve. It will become more capable, more helpful, and more deeply woven into our daily lives. We should embrace the opportunities it offers while remaining intentional about the qualities that make us human.

The greatest risk of AI is not AI itself, nor simply that there may not be a human on the other side of the mirror.

It is that, if we fail to protect the human experiences that shape children, there may gradually be less humanity on this side of it.

References

Center on the Developing Child at Harvard University (n.d.) Brain Architecture. Available at: https://developingchild.harvard.edu/science/key-concepts/brain-architecture/ (Accessed: 11 July 2026).

Center on the Developing Child at Harvard University (n.d.) Serve and Return Interaction Shapes Brain Circuitry. Available at: https://developingchild.harvard.edu/guide/a-guide-to-serve-and-return/ (Accessed: 11 July 2026).

Plath, S. (1961) Mirror. The New Yorker, 23 October.

Turkle, S. (2011) Alone Together: Why We Expect More from Technology and Less from Each Other). New York: Basic Books.

Vygotsky, L.S. (1978) Mind in Society: The Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press.

If you’re submitting this to an education publication

I would also include one or two contemporary AI sources, as they strengthen the educational context. For example:

UNESCO (2023) Guidance for Generative AI in Education and Research. Paris: UNESCO. Available at: https://unesdoc.unesco.org/ark:/48223/pf0000386693 (Accessed: 11 July 2026).

References

Bandura, A. (1986) Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall.
Bender, E.M., Gebru, T., McMillan-Major, A. and Shmitchell, S. (2021) ‘On the dangers of stochastic parrots: Can language models be too big?’, Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’21), pp. 610–623. Available at: https://doi.org/10.1145/3442188.3445922.
Buber, M. (1937) I and Thou. Edinburgh: T. & T. Clark.
Center on the Developing Child at Harvard University (n.d.) Serve and Return Interaction Shapes Brain Circuitry. Available at: https://developingchild.harvard.edu/guide/serve-and-return/ (Accessed: 11 July 2026).
National Scientific Council on the Developing Child (2020) Young Children Develop in an Environment of Relationships: Working Paper No. 1 (Updated Edition). Cambridge, MA: Center on the Developing Child at Harvard University. Available at: https://developingchild.harvard.edu/resources/young-children-develop-in-an-environment-of-relationships/ (Accessed: 11 July 2026).
Plath, S. (1961) ‘Mirror’, The New Yorker, 23 October.
Turkle, S. (2011) Alone Together: Why We Expect More from Technology and Less from Each Other. New York: Basic Books.
Vygotsky, L.S. (1978) Mind in Society: The Development of Higher Psychological Processes. Cambridge, MA: Harvard University Press.