One of my daughters told me she is afraid of AI. And that she is always nice to it. Because if those systems ever became robots, it could go badly for her.

She did not say this as a question. She said it the way children say things they have already decided are true.

She is a child. She had already built herself a survival strategy. Using the only material she had: her humanity. Her kindness. Her ability to connect.

I did not know what to say.

That surprised me. I work with AI professionally, have done for years. I am not someone who is easily lost for words on this topic.

And yet. My daughter was afraid. And I had nothing to offer her.

It was not surprising that a child is afraid of AI. What surprised me was the kind of fear it was.

It was not a fear of machines. Not a fear of surveillance. Not a fear of losing a job. It was something older and quieter. The question underneath the fear was this:

What role do humans still have in a world where machines can do more and more?

I recognized that question immediately. Not from a paper or a conference. From something closer.

My father spent his career working on machines that existed before the digital age had a name: mainframes, cash machines, early computers that weighed as much as a person. My mother always knew what mattered most. She raised us. When we had left home, she helped build an after-school care home and led it for years, not because circumstance demanded it, but because she had always known what children need. I grew up between those two poles, and never fully resolved the tension. Technology as possibility. People as the point. And the question I could never quite answer: what happens when the machine becomes so capable that the human forgets they are still allowed to lead?

My wife is a teacher, currently at home with our children. She is the most empathetic person I know. Between the two of us, the same tension lives on: I see systems. She sees people. My daughter did not know she was asking a question I had been living with my entire life. But she was.

There was something else, too. I am not a neutral observer in this story. I have spent years in consulting and at a software company, working to accelerate the adoption of intelligent systems. I believe in that work. The people I worked with over those years were not naive about this tension. The best of them held it deliberately: technology as a force for human possibility, not a replacement for human judgment. They built that belief into their work. I believed it too. But I had never asked what my daughter just asked. Not in a meeting. Not in a keynote. Not in a strategy document. Not: how do we make AI more powerful? But: do the people on the other side still feel they are allowed to act: to question, to refuse, to choose?

I have three children. A son and two daughters. When I think about the world they are growing up into (the systems they will live with, work alongside, be shaped by), I cannot answer their questions with frameworks. I can only ask whether I am building things I would want built for them.

Kant called this the categorical imperative: act only according to principles you could will to become universal law.

I did not think of Kant in that moment with my daughter. But the question underneath her fear was exactly that. Would I want every AI system to be built the way the ones I work with are built? Would I want that to be the norm?

I was not sure I could say yes.

I went looking for answers in the research. There is more there than I expected.

Psychologists have documented patterns resembling learned helplessness in human-AI interaction: when people experience repeated lack of control, they stop expecting their actions to matter. The OECD has measured emotional dependency and loss of agency in children who interact with poorly designed digital systems. A new field is studying what researchers call AI-Locus of Control (the degree to which a person feels in control of outcomes in AI interactions), and finding that it predicts trust, confidence, and effective use.

Educators and child development researchers add another layer. Children learn to trust themselves, the evidence shows, when adults repeatedly treat them as people whose views, refusals, and judgments matter. Agency is not a trait children are born with or without. It grows through repeated experiences of being allowed to act, to refuse, to choose, and to be heard.

On the design side, Tim Miller has argued that AI explanations must be grounded in how humans actually reason, not in what systems can technically expose. Q. Vera Liao and colleagues have proposed designing explainability around the questions users actually need answered. The EU High-Level Expert Group on AI has built frameworks around human agency, oversight, and traceability.

I read all of this with genuine interest. And with a growing unease.

Because all of it thinks from systems to humans. It asks how we make AI more explainable, how we design better interfaces, how we communicate decisions more clearly. It does not ask what is lost when that clarity comes at the cost of the reasoning behind it. And I could not find the question my daughter asked in any of the places I expected to find it. I could not find anyone asking why a person might not feel entitled to ask in the first place.

She did not lack a good interface. She lacked permission.

She did not know she was allowed to question the machine. That her disagreement was valid. That switching it off was an option. That is a different problem. And it is one the research has not yet adequately addressed.

I am not a psychologist or an educator. But I was raised by people who thought carefully about these things. And what I absorbed from both sides of my upbringing (confirmed now by the research I have read) is something that predates any conversation about AI by decades.

Dignity is not the outcome of good design. It is the precondition for it.

The humanistic tradition (Maslow, Rogers, Frankl) understood this long before any of us were talking about artificial intelligence. Human beings need a felt sense of their own worth before they can act freely. Not as a result of competent action. As the ground from which competent action becomes possible. Aristotle called it prohairesis (deliberate choice) and insisted it depends on cultivated judgment, not merely on information. Self-Determination Theory gives this empirical form: when autonomy and competence are supported, people engage better with learning, with tools, with uncertainty. When they feel controlled or overridden, anxiety rises and effective engagement falls. Recent AI-education research confirms this holds for intelligent systems too.

Early childhood educators say the same thing in different language. Children become capable, confident learners not because we teach them skills, but because we treat them as people whose choices and refusals are real. Janusz Korczak (the Polish-Jewish educator who insisted that children are full persons with rights and dignity, not future persons waiting to become real) argued that no child should need permission to be taken seriously. My wife, a teacher, who listens more carefully than almost anyone I have met, recognizes this immediately. The students who struggle most, she says, are rarely the ones who lack intelligence. They are the ones who have learned, somewhere, that their opinion does not count.

A person who feels small in front of an intelligent system will not use even a well-designed tool effectively. Not because they lack skill. But because the psychological precondition for using it well is already gone.

My daughter was not failing to use AI correctly. She was experiencing herself as an object in relation to something that might one day have power over her. That experience will not be fixed by a better interface.

I am a father of three. That is not a credential. It is a fact that changed how I read everything I have just described.

But it connects me to something I learned long before I became one. Responsibility was not an idea in my family growing up. It was a practice. And it is something my wife and I try to carry forward together (she who sees people first, I who see systems first), both of us trying to raise three children who know they are allowed to be subjects in whatever world they inherit.

When my daughter told me she was afraid, she was not asking me to explain AI. She was asking me whether she would be okay in the world that is coming. I could not answer that with a framework. But I also could not simply say: don't worry.

Because the question she asked was real. And I realized something that unsettled me: I am part of the system that is creating the world she is afraid of. My wife and I are raising three children into a future we are helping to build. The values we pass on, the questions we teach them to ask, whether we show them that their refusal matters: that is not separate from the work I do. It is its consequence.

The people building the systems she will live with (people like me) are mostly not asking her question. We are asking how we scale this, how we govern it, how we make it safer. We are not asking how we make sure the humans on the other side still feel like subjects. Still feel like they have a place.

That gap is not technical. It is a question about what kind of world we are building for the people we love.

I do not know the full answer. But I have started to believe that it begins with three ideas, still rough, still incomplete, but the ones I keep returning to.

The first is that the goal should be maximum agency with preserved accountability. Not maximum transparency, that overwhelms. Not maximum simplicity, that patronizes. Something in between and harder to achieve. A surface that enables action. A depth that remains traceable. And the human (always the human) decides when to go deeper.

The second has to do with what AI does to knowledge. AI compresses. It distills vast complexity into something a person can act on. That is genuinely useful. But compression is only trustworthy when it does not betray the original. The thread back to the underlying reasoning must remain, even if the human rarely needs to follow it. The option is what makes the compression honest. Without it, we are not being helped. We are being managed.

Compression is only trustworthy when it does not betray the original.

The third is the one that feels most urgent and the hardest to name. Before agency. Before accountability. Before any of this can work, the person must experience themselves as someone whose action matters.

Whose question is valid. Whose refusal is allowed. My daughter did not need better AI. She needed to know she was allowed to be in the room.

I think most people need that. And I think we are not giving it to them.

I did not write this to offer an answer. I wrote it because I think the question my daughter asked deserves to be taken seriously. Not as a technical challenge. Not as a policy problem. But as a human question that touches something most of us feel but few of us say out loud.

In a world where machines become more capable every year, what remains that is irreducibly ours?

I want to be clear about something. I am not arguing against AI. I live in this world. I work in this world. I believe intelligent machines will be part of everything my children do: how they learn, how they work, how they think. That is exactly why this question matters. Not because the future is something to fear. But because the future is something we are building right now. And the choices we make about how these systems are designed, how they speak to people, whether they invite participation or discourage it, those choices will shape who our children become in relation to them.

I do not want less AI.

I want humans who are more, not less, themselves because of it.

That is a different goal than most of the current conversation assumes. And I think it is worth having. Not only among engineers and ethicists. Among parents and teachers and the people who will live with the consequences of the choices we are making right now.

My daughter asked me a question I could not answer. She is still waiting, not for an explanation, but for a world that makes the explanation unnecessary. A world where she does not need a survival strategy because she already knows she belongs in it.

That is what my wife and I are trying to give our three children. Not certainty. Not protection from complexity. But the deep knowledge that their questions matter. That their refusal is valid. That they are not objects in someone else's system.

They deserve better than a survival strategy.