Not everyone who teaches is a teacher — the problem with robot tutors

I recently spent two days in a stakeholder engagement on the draft of the new Minimum Requirements for Teacher Education Qualification, the policy that will shape how our universities prepare teachers for years to come. Around the table were representatives from public and private universities, teacher unions, the Department of Higher Education and Training, the Department of Basic Education, the Council on Higher Education, the South African Council for Educators, the Joint Education Trust and members of the Education Deans Forum, among others.

For two days we worked through the draft, asking what should be included, what should be excluded, and what it had failed to anticipate. Each stakeholder made their case and perceived it as serious and necessary work. Yet, towards the end of the second day, in the closing discussion, someone made a statement that has stayed with me: “Not everyone who teaches is a teacher.”

The statement arrived at a crucial moment, because our country is about to test its truth on a national scale. The same policy must now prepare teachers for schools in which artificial intelligence is already present, and in which something more dramatic is beginning to arrive.

Universities have spent the past few years struggling to understand what AI means for teaching, learning and assessment. Our first response was largely to police its use by applying plagiarism-detection software, surveillance tools and attempts to distinguish the student’s work from AI tools. When these measures failed to resolve the deeper problem, many universities quietly settled into the position that students will use AI anyway, so perhaps we should simply learn to live with it.

And now, into this unresolved confusion, comes the next development: humanoid robots are beginning to enter South African classrooms as tutors. The question, then, is no longer simply whether robots can teach. It is whether teaching, by itself, is what makes someone – or something – a teacher.

Consider the road we have travelled to get here. First, we trained teachers to use whiteboards. Then came the flood of apps, including social media platforms such as Facebook and WhatsApp. Then Google Classroom and its suite of tools. Then the large language models. And now, it seems, we are to train teachers to operate humanoid robots. Somewhere in that long march of tools, a question has quietly gone missing, and it may be the most important one of all: what has happened to good teaching?

I am no longer sure we know what good teaching is, or even what makes a good teacher. And if we have forgotten that, we are in no position to decide what a robot should replace. Let me turn to the robots, because they are real, and they are coming. South Africa has opened a factory in Durban to manufacture a humanoid AI tutor called IRIS, created by former teacher Thando Gumede. IRIS can teach across subjects from Grade R to tertiary level in all 11 official languages, offering personalised lessons and instant feedback. The stated ambition is to place one in each of the country’s roughly 23,000 schools.

I want to be careful and fair here, because it would be easy and wrong to sneer at this. IRIS is a uniquely South African invention, built by a young Black woman from rural Hluhluwe who taught herself to code because she could not afford developers, and who spent years being told her idea would fail. Her motivation, she says, is not profit. She watched teachers struggle under impossible workloads and rural children go without the support that wealthier schools take for granted. She built IRIS not to replace teachers, but to help them. And the problem she is responding to is real. It is one we cannot simply wish away. South African teachers are carrying burdens that no technological optimism can make disappear.

But to understand what a robot in a classroom would mean, I did something simple. I spoke to two teachers, in two very different schools. The first teaches in a school marked by high levels of substance abuse, the kind of underresourced, high-need school our system has always failed. She did not hesitate to say smilingly: “I don’t want to be funny, but can this robot defend itself? These kids will tear it apart.” She continued: “Did they first test it in school like ours, otherwise it will just be more work, having to guard it during our breaks.” Then, with the weary generosity I have come to associate with teachers in the hardest schools, she added: “But let’s hope it can add value.” The second teacher works in a quintile 5 school, well-resourced and stable. Her response could not have been more different. “It’s a great idea,” she said, “and we look forward to using it effectively. It might help ease our work.”

Two teachers, two worlds, the same robot. And in the gap between their answers lies much of what worries me. The teacher in the wealthy school sees a helpful tool that will ease an already manageable load. The teacher in the poor school sees a fragile object that will be vandalised and become one more thing to police, in an environment already stretched past breaking point. The same robot means opposite things in the two halves of our divided education system. That difference should make us pause before we celebrate. Because it tells us, before we have argued anything else, who this technology will serve and who it may simply be imposed upon.

So let us ask the question from that policy meeting directly: Is the humanoid tutor robot that teaches, a teacher? To answer, we must recover the thing we seem to have forgotten. If teaching is simply the delivery of content, the transfer of information from someone who has it to someone who does not, then a robot can teach. It may soon do so better than an exhausted human being with 60 children in front of her. It can explain a concept 100 different ways without losing patience, mark immediately, and never have a bad day. If that is all teaching is, then IRIS is not a threat. She is the answer to our problems.

But I do not believe that is what teaching is. Neither I suspect, do you, if you think back to the teachers who changed your life. We do not remember them primarily for the information they delivered, information we could probably have found in a textbook anyway. We remember them because they saw us. Someone noticed when we were struggling and stayed behind. Someone believed in us before we believed in ourselves. Good teaching has never been only about the transfer of content. It is a human relationship in which one person is fully present to another and helps them become someone they were not before.

That, I think, is what the sentence from that meeting was pointing towards. Many things can deliver content but that does not make them teachers. To teach, in the fullest sense, is to be present to a pupil as a person. It requires being awake to what is happening in front of you, to the hesitation in a child’s voice, the silence that means something is wrong, the moment when a pupil needs you to stay. It is this presence that a robot does not have.

A colleague in our faculty and I run a project in which we take our pre-service teachers into a children’s hospital each week to work with children who have been ill for months and cannot attend school. There is no syllabus there, no test, and often the child is too weak to hold a pencil. What our students learn at those bedsides is that the deepest thing a teacher does is sometimes simply to be present, to stay, to attend, to be a human being who is truly there for a frightened child. That is teaching stripped to its essence. And it is the one thing no robot can do, not because the task is too difficult, but because presence requires a living being for whom the child matters. IRIS can simulate warmth, but she cannot care. She can recognise a face, but she cannot be present to a child. She can teach in the thin sense of delivering content, but she cannot be a teacher in the fuller sense of accompanying the development of a human being.

This is why I want to ask a hard question about where IRIS could take us; not as an attack on Gumede’s intentions, which I admire, but as a warning we would be foolish to ignore. The two teachers have already given us the answer. Wealthy schools will keep their qualified human teachers and use the technology to assist them. It is the poor and rural schools, the ones that already cannot attract teachers, that will be handed robots instead. So, the very inequality IRIS is meant to close could widen into something far worse. The child from an affluent community will get a present human being while the poor child will get a robot. And we will call it equity, because both now have access to the technology. This is not a hypothetical, it is simply the logic our two-nation, bimodal education system has followed for a century, handing the poor the cheaper version and calling it fairness.

There is real force in the reply that a robot is better than nothing, and in too many of our schools nothing is the honest alternative, an empty post, a subject no one can teach. A child with no teacher may be helped by IRIS, and I will not pretend otherwise, but here is the danger. Once robots can be dropped into underresourced schools, the pressure to do the harder and costlier thing – to train, attract, support and keep human teachers where they are most needed – quietly evaporates. The robots let the state off the hook. What was offered as a bridge to equity becomes the permanent reason the poor are given a substitute for the thing the rich still enjoy. The stopgap becomes the settlement, and a generation of poor children grows up taught by robots while being told it is the same as being taught.

No, it is not the same and this is where the sentence from that policy meeting finally comes to rest. As we write the rules that will govern how this country prepares its teachers, and as we decide what place these robots will hold in our schools, we must hold on to the distinction we are in danger of losing. Not everyone who teaches is a teacher. A robot can teach but it cannot be a teacher. AI is here and, used wisely, it can lift real burdens from overstretched teachers and give struggling pupils real help, and we should welcome that. But we must never confuse the delivery of content with the cognitive development of a child, or a clever humanoid robot with a present human being. If we make that mistake, it will not be the children of the comfortable who pay for it. It will be the poor, handed a beautifully engineered robot and told that it is a teacher. It is not, because teaching is presence, and presence is the one thing we must refuse, ever, to automate away. DM