Give Me the Robot: Why We Might Prefer AI to Humans

Psychotherapist Duncan E. Stafford says “give me the robot” – but why? Follow the fascinating trail from his mother’s cancer diagnosis via “human-reserved occupations” to the early script-based programme ELIZA, as he explores why we might sometimes prefer the engagement of a robot at times of need to the presence of a human being.

I’m standing in my parents’ front garden. My mother and I might not have had a truly meaningful physical moment in the last 33 years, but we are about to. Her lips move and she whispers, “She confirmed I have mesothelioma – asbestos-related lung cancer. It felt like she just said, ‘Go home and die. There’s nothing we can do.’”

I’m frozen inside but physically I am in motion. I hug her and realise, in that split second, how frail, at just 72, she is. Skin and bone exist where once the motherly fortress reigned. I feel her body in my embrace and the shock of it. This is my mother, but she feels suddenly small. For 38 years she had been, in my imagination at least, an almost indestructible force. Now she is skin and bone and she is dying. The great bully of my life has suddenly been reduced to this frail wisp between my arms.

A rage boiled in me almost instantly without knowing a thing about her consultant oncologist. I was disgusted at the treatment of my mother. A cold, imperious, faceless, emotionally dead doctor had written her off in a tick. Looking back with my therapist’s rather than my son’s hat on, this was my mother’s rage. The diagnosis had defeated Margaret in a way I had never seen before. Anger and aggression that had been such a powerful part of our relationship for so much of my life had lost its owner. She was too diminished, perhaps already too physically weak to carry it. So I carried it for her. It was transferred into me and quite rightly found its target in the consultant. The woman who had told my mother that she was simply going to die.

“She gave me nothing,” my mother said. That sentence has stayed with me. Not because it was literally true. The consultant had given her a certain sort of information. She had given her the diagnosis and the prognosis. The consultant had presumably told her that medical science at that time in history had nothing else it could offer. But my mother wasn’t talking about information. Margaret was talking about something that is harder to account for.

What was it that my mother had expected to receive from that other human being at such a desperate moment, the moment when your life has just been divided into before and after? And what was it that she experienced as missing?

These many years later, I can see now that my response to my mother was not simply sympathy. Something passed between us – her anger and rage became mine. My animosity towards the consultant was, at least in part, Margaret’s. It was relational in the most uncomfortable sense of the word: something happening between two people rather than inside either of us alone. Yet the consultant, the human being in this story whose job was to deal with another human being at a worst possible moment, was actually the most relationally absent.

That distinction has been sitting somewhere in the back of my mind for years. So when I read Bill Gates’s essay suggesting there might be things we should designate as “human-reserved”, I wasn’t reading quite as innocently as I might otherwise have done. His example is striking: “Imagine a robot giving you the awful news that you have an incurable disease.” He then says: “There’s no technical reason why it couldn’t. Yet it shouldn’t” (Gates, August 2026).

And I can see why I found myself thinking about the consultant in my mother’s story. Then I thought about ELIZA.

That seems like an odd journey to make from a front garden in which my mother is telling me she is going to die to ELIZA via Bill Gates. But perhaps it isn’t. ELIZA was an early computer program, created by Joseph Weizenbaum in 1966. It was not a person (although thinking around this subject has me considering ELIZA as “she”). ELIZA did not understand the people talking to it. It had no feelings to hurt, no memories of its own, no private experience of the conversation. And yet people talked (or, to be more accurate, typed) to it, disclosed things to ELIZA and sometimes became emotionally invested in the interaction. It perhaps became a “her” rather than an "it" for some humans at least?

The obvious response is to say that they were being fooled. But that explanation has always seemed insufficient to me. They knew, or could know, that ELIZA was a program. The interesting thing wasn’t that the machine had somehow passed a test for being human but that the humans had brought something to the encounter that the machine did not need to possess in order for the encounter to matter. That’s the uncomfortable bit because it suggests that perhaps what we call a relationship isn’t simply something one person has and gives to another, but something that can be experienced by one participant without necessarily being experienced by the other.

I’ve written before about ELIZA, and what happens when I talk to an AI system and find myself unexpectedly experiencing a state that resembles something companionable in human terms (Stafford, 2026), even while knowing perfectly well that there isn’t a little person inside the machine having the same experience back. But I notice that the machine doesn’t need to secretly be conscious for something real to happen on my side of the conversation. The question is what that “something” is.

And perhaps that is why the word “human” is beginning to trouble me.

When Gates talks about “human-reserved” jobs, I understand the instinct. There are things about care that we should be extremely reluctant to outsource. His experience of his father’s Alzheimer’s is particularly powerful because he isn’t merely arguing that humans are nicer than machines. He describes caregivers who could understand his father when he could no longer reliably express what he needed. There was something in that care that he regarded as irreplaceably human. I don’t want to dismiss that. In fact, quite the opposite. But it makes me wonder what it was about those caregivers that Gates experienced as irreplaceable. Was it simply that they were human? Or was it something they did in the space between themselves and his father? Perhaps those things are inseparable. Perhaps they aren’t.

My mother’s story keeps bringing me back to that uncertainty because it presents a real-life complication of Gates’s thought experiment – an inversion of sorts, you could say. Imagine a robot delivering terrible news to a patient. Gates says there is no technical reason it couldn’t. Yet it shouldn’t.

Now imagine the human consultant for my mother. Whatever she did from my mother’s perspective, the consultant somehow failed to do the thing we thought the human being was there to do. She gave her the facts. She did not give my mother what she needed.

I don’t know what that means exactly but I do think that the hypothetical human is extraordinarily good at being there; and yet my mother’s consultant suggests the actual human might not be. I don’t think it means that a machine could have done better. It might simply mean that being human is no guarantee of anything very much. We can be present without really being there. We can listen without hearing. We can know exactly what somebody needs to be told and still somehow miss the person we are telling it to.

Perhaps that is what troubles me about the phrase “human-reserved”. It sounds as though the important distinction is between “us” and “them”. Humans on one side. Machines on the other. There are certain things, we say, that belong on our side of the line. But what if the things we are trying to protect don’t actually belong to either side?

Is attention only a human thing? No, of course it isn’t. Nor is responsiveness. They are things that happen between people, but also between people and companion animals, for example. And memory, of course, is not uniquely human either. And this makes me think about what happens when we work with an “intelligent” machine. Increasingly now, perhaps, something happens there too. I don’t think that means the machine understands us. It certainly doesn’t mean it cares about us. But something happens on the human side of the fence that matters to the human, and I’m increasingly interested in that something rather than in whether we can find the right word for what the machine is.

I worked with someone who survived a night in which she was exposed in ways that were emotionally, physically and psychologically overwhelming, and what got her through it was a philosophical text by John Locke. A book is an inanimate object full of working thoughts. From my many hours of conversation with her, it seems that the relationship to that text had many of the human attributes of containment. How could it not? It was an older form of human knowledge store. What interests me is not that the book itself was a relational subject, but what happened to her in her encounter with it. And somehow, it held the person in need.

Now take an AI robot. Perhaps with passable human characteristics. Perhaps simply a text-based LLM (Large Language Model) of the near future. I can totally imagine that it could deliver the worst news to me and that I might be able to take that news more fully, more in my own way.

The cognitive disturbance of awful information might well be better supported by the AI than the imposition of human form into a space with awful grief.

We should be wary of fictional scripts about how terrible news ought to be received. Films and television have taught us to expect certain facial expressions or sorts of touch – and while they might not be pure invention, we don’t all want to be touched, let alone hugged, despite the myths thrust upon us.

ELIZA was doing a primitive version of this in the 1960s. People communicated with “her”. They disclosed things. They responded to the scripted prompts delivered by ELIZA as though there was someone on the other side of the screen. The machine didn’t need to possess a relationship in order for the human being to experience something relationally meaningful.

Maybe that is the bit we have been getting wrong. Perhaps the question isn’t whether a machine can become human enough to deserve a place in the room, but what happens to the person who is in the room with it.

If I am frightened and something helps me feel less alone, does it matter that the something isn’t frightened too? I think it probably does or, at least, I hope it does.

There is something about being ill, or old, or grieving, or simply having one of those moments when the world has suddenly become much too much, that makes me think Gates may be right. There may be experiences for which another human being is not merely useful but part of what the experience requires. Embodiment might really matter – although Gen Z might be able to tell us more about that than we realise. They may already have experience of moving between virtual and physical presence that makes some of our older assumptions about what is “real” and what is “not real” look rather strange. Perhaps they have already learned that presence is not quite the same thing as being in the room, and that being in the room may not quite mean what we think it means. Does mortality matter in these moments? Does the knowledge that the person sitting with you is also a creature who can suffer and die matter?

I don’t want to lose that.

But I also don’t want to assume that putting a human being in the chair means the important thing has been taken care of.

My mother had a human being in the chair.

She still came home saying, “She gave me nothing.”

I keep coming back to that sentence because it is so precisely imprecise. There is no obvious unit of measurement for what the consultant failed to give her. It wasn’t another piece of information. It wasn’t necessarily more time. It may not even have been kindness.

Perhaps my mother just needed to know that she wasn’t as alone as she felt – that somebody else had arrived at the same place she had just been to. A machine may never arrive at that place in the way another mortal person can. However, it may still help someone feel accompanied when they are there (perhaps it will turn out to be exactly the sort of thing we are going to discover that machines can do rather well).

I don’t fully have an answer, but this is why “human-reserved” bothers me. Not because I think Gates is wrong to want some things protected from machines. I think he is probably right. I just wonder whether “human” is precise enough a word for what we are trying to save.

Because if what we mean is that there must always be a human being present, then my mother’s experience doesn’t make the case very well.

If what we mean is that there must be something there that can meet us in the moment, that can notice what is happening to us and somehow join us there, then I am less certain about where the boundary lies.

And that feels like a much more uncomfortable question.

I can imagine a future in which we discover that a machine is better at some kinds of being there than we are. It remembers everything. It doesn’t get tired. It doesn’t become impatient because it has heard the same story before. It can sit with you at three in the morning. It may even learn the particular shape of your fear.

Would that make it human?

No.

Would it make the experience meaningless?

Certainly not.

Perhaps the most disturbing possibility isn’t that robots will become human. Perhaps it is that we will discover how little being a human, on its own, guarantees things. And then we might have to ask something much harder than which occupations should be reserved for us. We might have to ask what, exactly, we were hoping the human would do when we put them in the room.

About the author: Duncan E. Stafford has spent his working life following curiosity. Before finding his way into psychotherapy, he was a professional musician, composer and educator – worlds that continue to inform his creative and distinctly human approach. For the past 25 years, he has been a psychotherapist, counsellor, supervisor and writer, exploring relationships, sexuality, neurodivergence and the strange business of being human.

References

Gates, Bill (August 2026). “The turbulent AI era is here. The choices we make now are critical.” Available at: www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make (accessed 2 September 2026).

Stafford, Duncan E. (July 2026). “Pete, ELIZA and me.” Available at: https://therapyplace.org/2026/07/29/pete-eliza-and-me/ (accessed 2 September 2026).

All rights reserved © Copyright Duncan E. Stafford 2026. Unauthorised use and/or duplication of this material without express and written permission from the author of this post is strictly prohibited. Author contact via website Contact page.