A person does not only need a response. A person needs to be recognized.
In South Africa, where the wounds of apartheid still shape who receives care, recognition, and dignity, a scholar warns that generative AI may quietly inscribe a new hierarchy — not through law, but through the substitution of machines for human presence in the places where personhood is affirmed. Drawing on ubuntu, the philosophy that holds a person becomes themselves only through genuine encounter with others, she argues that routing the already underserved toward chatbots while the privileged retain human attention is not merely an efficiency choice, but a moral one. The question South Africa now faces is whether technology will widen the circle of human recognition or simply automate its exclusion.
- Generative AI is spreading rapidly into customer service, healthcare, and education — the very spaces where vulnerable people most need to feel seen and heard by another human being.
- A scholar has named the emerging risk 'relational apartheid,' a term that deliberately echoes South Africa's history of determining who counted as fully human through systems of separation.
- The tension is not about whether AI gives correct answers, but whether a machine can ever offer the mutual vulnerability, accountability, and recognition that constitute genuine human dignity.
- Tiered access is already taking shape: wealthier users escalate to human agents while underserved populations are managed by bots, threatening to encode old inequalities in new digital architecture.
- Proposals for ubuntu-aligned AI — culturally sensitive, community-designed systems — offer partial promise, but scholars caution that ubuntu cannot be programmed, only supported by human relationships it is designed to protect.
A machine can apologize, remember, and adjust its tone to match your mood. For someone lonely or frustrated, a chatbot can feel surprisingly like a person. But feeling like a relationship and being one are not the same thing.
This distinction carries particular weight in South Africa, where a scholar of technology and society warns that the rapid spread of generative AI risks creating what she calls a 'relational apartheid.' The term is deliberate. South Africa's legal apartheid determined who was seen, heard, and recognized as fully human. That legacy persists in unequal access to education, healthcare, and public services. Now, a new form of inequality may take shape not through law but through the quiet substitution of machines for human presence in the places where dignity is affirmed.
Her argument rests on ubuntu — the South African philosophy that holds 'a person is a person through other persons.' Dignity, in this view, is not individual and isolated but relational, formed through mutual recognition, care, and shared life. A chatbot can simulate responsiveness, but it cannot share in your life, be vulnerable in return, or be held accountable as a person. Human relationships are difficult precisely because they involve mutual risk. A machine offers responsiveness without that resistance — always available, endlessly polite, easily reset. That convenience is attractive, but it may train us to expect connection without the labor that makes it real.
The danger is most visible in how AI is already deployed across customer service, healthcare, and education. As organizations automate their front lines, a tiered system emerges: some people receive human attention while others are managed by systems. In a country the World Bank describes as among the most unequal in the world, those already underserved risk being increasingly spoken to by machines rather than people — not because of explicit policy, but because efficiency has become the primary measure of progress.
There is talk of building ubuntu-aligned AI, systems sensitive to African languages and communal values. This has merit. But ubuntu cannot simply be programmed. It depends on shared life, mutual vulnerability, and accountability that current AI systems cannot live the way people do. The path forward requires that AI be presented clearly as a tool, not a companion — supporting human professionals in sensitive settings rather than replacing them, and ensuring that the measure of a good system is not only what it answers, but who it leaves room for.
A machine can apologize. It can remember what you told it yesterday. It can adjust its tone to match your mood, offer encouragement when you sound discouraged, and respond with what feels like genuine warmth. For someone lonely, frustrated, or simply seeking reassurance, a chatbot can feel surprisingly like a person on the other end of the conversation. But feeling like a relationship and being one are not the same thing.
This distinction matters urgently in South Africa, where a scholar of technology and society is now warning that the rapid spread of generative AI—the large language models that power chatbots and digital assistants—risks creating what she calls a "relational apartheid." The term is deliberate. South Africa's history of legal apartheid did not merely separate people by law; it determined who was seen, heard, and recognized as fully human. That legacy persists today in unequal access to education, health care, work, technology, and public services. Now, a new form of inequality could take shape not through explicit law but through the quiet substitution of machines for human presence in the places where dignity is affirmed.
The scholar's argument rests on ubuntu, a South African philosophy of personhood that sees human dignity not as something individual and isolated but as something relational—formed through mutual recognition, care, responsibility, and shared life. Ubuntu holds that "a person is a person through other persons." This is not merely poetic. It describes how people actually become themselves: through encounters with others who see them, know them, and hold them accountable. A chatbot cannot do this. It can simulate responsiveness, but it cannot share in your life. It cannot be vulnerable in return. It cannot be held accountable as a person. It cannot forgive, be forgiven, carry a moral burden, or be transformed by knowing you. Human relationships are difficult precisely because they involve mutual risk—disappointment, misunderstanding, the hard work of repair. A machine offers responsiveness without that resistance. It is always available, endlessly polite, easily reset. That convenience is attractive. But it may also train us to expect relationships without the labor that makes them real.
The danger emerges most clearly in how AI is already being deployed in customer service, health care, education, and social support. As organizations automate their front lines, customers are increasingly routed through chatbots before they can reach a human being. This reduces costs and may speed simple queries. But it creates a tiered system: some people receive human attention; others are managed by systems. Salesforce, one of the world's largest providers of customer service software, has reported that AI agents now handle a growing share of customer interactions. The company has also reduced thousands of support roles in recent years. Something similar could unfold in health care, education, and social support. Where human professionals are already scarce—as they are in many parts of South Africa—AI counseling tools, tutoring systems, and advice bots may appear to offer a practical solution. They may help in some cases. But they also risk normalizing a situation in which those already underserved are increasingly spoken to by machines rather than people.
In a society the World Bank describes as among the most unequal in the world, old differences in income, education, language, geography, and institutional power could reappear in new digital forms. The problem is not whether machines give accurate answers. The deeper problem is that some people may be denied the kind of encounter through which dignity is affirmed. A person does not only need a response. A person often needs to be recognized by another person.
There is talk of building ubuntu-aligned AI—systems more sensitive to African languages, local histories, and communal values, designed with community involvement rather than imposed by distant corporations. This has value. But ubuntu cannot simply be programmed into a machine. It is not a set of polite phrases or cultural preferences. It is a way of thinking about persons in relationship, and it depends on shared life, mutual vulnerability, and accountability. A system can be designed to support these values, but current AI systems cannot live them the way people do.
This distinction matters for policy and design. AI systems should be presented clearly as tools, not companions. They should not blur the line between simulated care and real care, especially when used by children, older people, patients, or socially isolated users. In sensitive settings, AI should support human professionals rather than replace them. South Africa may use AI to improve public services, widen access to knowledge, and support overburdened institutions. But it may also deepen the distance between people if efficiency becomes the main measure of progress.
Notable Quotes
A person is a person through other persons— Ubuntu philosophy, cited by the scholar
The appearance of a relationship is not the same as a relationship— The scholar's analysis of AI-mediated interactions
The Hearth Conversation Another angle on the story
When you say "relational apartheid," you're not comparing it to the legal system of the past. What exactly do you mean?
I mean a pattern where access to meaningful human engagement becomes unequally distributed. Some people are met by persons. Others are managed by systems. It's not law. It's how technology and economics combine to create the same outcome—some people are recognized as fully human, others are not.
But AI can be helpful. It can teach, inform, support. Why is that a problem?
It's not a problem in itself. The problem is substitution. A chatbot can help you find information. But if it replaces the teacher who knows you, the counselor who can be held accountable, the doctor who shares the weight of your suffering—then something essential is lost. Dignity requires recognition from another person, not just a response from a machine.
You mention ubuntu. Why is that philosophy specifically important here?
Because ubuntu says personhood is relational. You become yourself through others. In South Africa, that's not abstract—it's how people have survived and resisted. If machines replace those relationships, you don't just lose convenience. You lose the ground on which dignity stands.
Can't we design AI to respect ubuntu values?
We can design systems that support human professionals, that are sensitive to local languages and histories. But ubuntu can't be programmed. It requires mutual vulnerability, accountability, shared life. A machine can simulate care. It can't live it.
So what should happen?
Be clear about what AI is. Use it to support human professionals, not replace them. In health care, education, social support—the places where recognition matters most—keep the human at the center. Otherwise, you risk normalizing a new form of the old inequality.