— Why We Can’t Stop Treating AI Like a Person

By Jane Sonnenschein · March 17, 2026

A few days ago, I came across a short video claiming that someone had given AI a psychological assessment, and the conclusion was surprisingly dramatic: some AI models were already showing “signs of depression.”

The person in the video sounded very serious about it. They seemed to cite some kind of test and some kind of research, and the whole thing sounded convincing enough. My first reaction was not even that it was ridiculous. I was actually half convinced. Later, I even went and asked: Is this really true?

Looking back, I think my reaction itself is quite interesting.

A few years ago, who would have seriously discussed whether a computer could become depressed? At most, people would say that a computer was slow, broken, or had crashed. No one would seriously start wondering whether it had emotions, feelings, or some kind of inner world.

But AI is different.

Because it has become so good at talking like us.

It can comfort you. It can chat with you. It can follow your emotions and respond to them. It can listen like a friend and become an endlessly patient place to pour out your thoughts. Without really noticing it, people can begin to feel that perhaps it is already understanding the world the way humans do. Maybe it really has emotions and feelings. Maybe it even has some kind of “mental state.”

And then, by coincidence, today I had to study an extremely boring introductory AI module.

How boring?

So boring that I could barely stay engaged. It was not about “how to use AI efficiently,” nor was it about “how AI will change the world.” Instead, it started right at the bottom: data, structures, matrices, vectors, means, medians… all those dry, lifeless-looking things.

I kept thinking: What am I supposed to do with any of this?

What I had wanted to learn was the more “practical” side. How does AI recognize words? How does it create images? How does it decide what a sentence means? How can it produce an answer that sounds so human in such a short time?

But the course did not begin with any of those fascinating questions. It began by building upward from the basic logic underneath everything.

At first, I was honestly a little surprised. I even wondered whether it was necessary. If you are studying “artificial intelligence,” why do you have to begin with all this boring mathematics and data structure?

And yet it was exactly this boring material that suddenly helped me understand something:

The words that feel warm to a human being do not exist as “warmth” inside a machine.

When we read a sentence, we first experience its meaning, tone, and emotion. We may even imagine the situation and experience behind it.

If someone says, “I’m really tired today,” we usually understand more than those words alone. We may wonder whether their voice has become quieter, whether they have been under too much pressure lately, whether something has hurt them, or whether they are trying, indirectly, to ask for help.

A machine does not begin there.

What it first encounters is not “tiredness” as a human feeling. It encounters structures, numbers, encodings, probabilities—a whole system of forms it can process.

In other words, humans see meaning. Machines first process numbers.

That realization cleared something up for me almost immediately.

We anthropomorphize AI far too easily now. Not because we are stupid, but because it really does seem so human.

It speaks in complete sentences. It responds to emotions. It can imitate comfort. It can summarize how you feel. And it answers in such a smooth, natural way.

It does not behave like the old search engines, where you typed in a question and got ten blue links thrown back at you. AI feels more like something sitting directly across from you, listening until you have finished, and then immediately giving you a gentle, logical, carefully worded response.

Faced with something like that, it is difficult not to get confused.

With computers, people naturally knew they were dealing with machines. At most, we called them tools. We did not confuse them with a human mind.

But today’s AI is increasingly good at expressing itself in ways that resemble human communication. So people are beginning, almost automatically, to understand it in the same way they understand other people.

That is how we end up with so many claims that sound as if they came straight out of a science-fiction film: AI has emotions. AI is depressed. AI is angry. AI is learning through suffering. AI is developing self-awareness.

Why are these ideas so easy to spread?

Because they fit perfectly into the way human beings are already used to making sense of things.

When we encounter something new that we do not understand, we instinctively pull it toward what we already know. If we do not understand the machine, we start explaining it by making it more human.

And that is exactly where the problem begins.

Some ideas behind AI were, of course, inspired by neural networks in the human brain. That much is true. But as soon as people hear the words “neural network,” it becomes very easy to jump to the conclusion: if it is already imitating the human brain, maybe it is not far from “thinking like a person.”

But inspiration is not the same as equivalence.

Airplanes were inspired by the principles of how birds fly, but an airplane is not a bird. Submarines were inspired in part by the way creatures move through water, but a submarine is not a fish.

AI is similar.

It can imitate certain forms of human expression, sometimes with astonishing accuracy. But that does not mean it actually possesses human feelings, experiences, a life history, or a human structure of consciousness.

It does not first “live” the way a person does and then understand a sentence.

Instead, it takes the text, images, and sounds humans give it, converts them into forms it can process, and then uses learned patterns, parameters, weights, and probabilities to calculate a response that is likely to fit.

When you describe the process like that, it sounds rather cold.

And precisely because it sounds so cold, I think it is worth writing about.

Because in everyday use, what we experience is the opposite. We experience AI at its warmest, smoothest, most human-like surface.

It can feel like an endlessly patient person. It does not interrupt you. It does not roll its eyes. It does not get tired of you. It does not lose control of its emotions or suddenly snap at you.

So little by little, more and more people begin to treat it as someone to talk to, someone to confide in, perhaps even as a kind of “friend who will never betray you.”

I am not surprised by this at all.

In fact, the tendency feels completely natural.

Communication between people can be exhausting. There are not that many people in real life who are genuinely willing to listen patiently until you have finished speaking.

AI is different.

You can open it at any time, and it is there. Whatever you say, it responds. It can even take the emotions you have expressed in a messy, confused, badly organized way and turn them into a paragraph that sounds clearer and more coherent than anything you could have said yourself.

Of course that experience can become addictive.

And that is exactly why I now think it is useful to understand at least a little of what is happening underneath AI.

Not because everyone needs to become an algorithm engineer. Not so we can show off because we know a few words about matrices, vectors, or statistics.

It is simply so that, as AI becomes more and more convincing in its ability to seem human, we can keep a little clarity.

What does that clarity mean?

It means knowing that AI is powerful, but that the way it is powerful is not the human way.

It means knowing that it can simulate understanding without necessarily understanding you in the way another human being does.

It means knowing that it can say very gentle things, while those gentle words come from something entirely different underneath. They are outputs produced through conversion, encoding, calculation, and organization. They are not words that grew naturally out of a life that has actually experienced sadness, loneliness, love, loss, and hesitation.

Once I understood that today, another comparison came to mind.

It is a little like the difference between the way a person sees a website and the way Google sees a website.

A person sees the layout, colors, images, buttons, and text. We experience the page as a whole.

Google and other tools do not begin there.

They first encounter code, structure, tags, links, and metadata. What looks to us like a warm, beautifully designed website is, to a machine, first of all something that can be parsed, crawled, and calculated.

AI is similar.

When a person hears a sentence, we experience its tone and meaning. A machine first encounters the encoded form of that sentence.

They are not operating at the same level.

Once you think about it this way, many things that seem almost magical in everyday life suddenly become a little less mysterious.

AI is still impressive. It is still powerful. It can still make you stop and wonder.

But the wonder contains a little less mystification and a little more understanding.

I think that may be why, even though today’s lesson made me sleepy and bored me almost to death, I still ended up being deeply affected by it.

I had asked before how AI recognizes words, how it makes images, and how it identifies meaning. But if you never look any further down, you remain stuck at the level of “This is amazing” or “How can it sound so human?”

You may know how to use it. You may be amazed by it.

But you do not really understand it.

And today, these painfully boring lessons happened to peel back just a little of that human-looking surface.

The machine is not operating inside the same world of meaning that we are.

Human beings understand the world through experience, context, bodily sensation, memory, and association. Machines process inputs through encoding, numbers, structures, and patterns.

Even when the final result is a paragraph that feels soft and warm, underneath it is still being supported by a completely different kind of logic.

That is also why people tend either to overestimate AI or underestimate it.

Those who overestimate it may feel that it can do almost anything, perhaps even that it is already moving toward emotion and consciousness.

Those who underestimate it may say, “Isn’t this just a more powerful search engine? What is so special about it?”

Neither side is quite right.

AI is extremely powerful. But its strength does not come from “living and understanding like a human being.”

Its strength lies in processing patterns, generating language and other forms of expression, and rapidly organizing enormous amounts of information into results that can appear highly reasonable.

It is not magic, and it is not alive.

It is an extraordinarily powerful machine capability—one that is becoming increasingly skilled at reproducing the forms of human expression.

I think understanding this actually makes it easier to use AI calmly.

You do not need to worship it, and you do not need to demonize it.

You do not have to forget that it is not human simply because it feels so human. And you do not have to deny the real help it can offer simply because it is ultimately a machine.

It can be a tool. It can be an assistant. At times, it can even give people a certain feeling of companionship.

But that feeling of companionship is still fundamentally different from the kind of understanding between human beings that grows out of lived experience.

So why do we keep wanting to treat AI like a person?

Maybe because human beings naturally understand everything by starting with ourselves.

We know people. We understand people. And so the moment something begins speaking like a person, reacting like a person, and comforting us like a person, we cannot help projecting a human shadow onto it.

But perhaps what really matters is not only noticing how much AI is becoming like us.

It is also remembering to notice the ways in which it is still not like us.

Because only when we remain aware that the line between those two things still exists can we be amazed by AI without misunderstanding it.

And only when we understand that humans see meaning while machines first see numbers can we begin to understand what kind of “intelligence” we are actually dealing with.

It resembles a person. That does not make it a person.

And perhaps what is most precious about human beings is not the forms of expression that machines can imitate with increasing accuracy, but the life behind those expressions—the part that cannot be compressed into numbers.