Artificial intelligence is usually discussed in terms of what it can do: how fast it can work, which jobs it might replace, or how it should be regulated. But the more fundamental question is what AI may change about us.

That was the focus of a recent United Nations convention on AI and Human Development that I attended in Xi’an, China.

The closed-door meeting brought together academics, technologists, artists, policymakers and UN representatives to consider a deceptively simple question: what does it mean to be human in a world increasingly shaped by artificial intelligence?

This is a crucial question for the global community because AI is no longer simply another new technology. It is entering education, work, health, culture and everyday relationships at remarkable speed.

The Xi’an meeting followed earlier discussions planned through UN Partnerships in Cairo and Mexico City and forms part of a wider UN process linked to the Pact for the Future, the 2027 SDG Summit and the Independent International Scientific Panel on Artificial Intelligence.

My contribution centred on four human-development problems.

Whose knowledge will AI recognise?

AI may widen access to information but it has biases about what knowledge is, and whose knowledge it includes.

A system can speak hundreds of languages and still rely on a relatively narrow understanding of ideas such as wellbeing, autonomy, family – even what counts as legitimate knowledge. Multilingual AI is not necessarily epistemically plural, meaning it privileges a few ways of knowing and excludes or neglects others.

This matters especially for Indigenous peoples and smaller societies. Concepts such as Country in Indigenous Australian traditions or Vanua in parts of the Pacific are not simply facts about land. They describe lived relationships among people, ancestors, place, spirituality, obligation and community.  

This is partly reflected in the Australian Framework for Artificial Intelligence in Higher Education that explicitly affirms Indigenous peoples' rights to protect and manage their cultural heritage and knowledge.

Turning these relationships into data may preserve information while changing the knowledge itself. There is also the thorny question that some forms of Indigenous knowledges are not necessarily translatable.

This also matters geopolitically. Countries such as Timor-Leste may become enthusiastic users of AI while having little control over the infrastructure, models, or datasets through which AI interprets their societies. AI becomes a process of socialisation but one divorced from the actual society itself.

Access can therefore increase at the same time as dependency. The question is not simply whether people can use AI, but whether societies can remain authors and co-controllers of the knowledge systems shaping their futures. This has become a distinct focus for AI use in higher education teaching and research to affirm Indigenous peoples right to protect and manage their cultural heritage.

Can a machine really recognise us?

Second, AI raises a problem of recognition.

Human beings are not isolated information processors. We develop our identities and autonomy through relationships with others. Recognition is part of being treated as a person whose experiences, needs, and value matter.

AI can simulate affirmation and even empathy and partnership. But there is a difference between receiving a convincing response from a system and being recognised by another person who is themselves part of reciprocal social relations.

That distinction becomes especially important in education, care, and mental health. The question is not whether AI can provide useful support – it clearly can – but whether efficiency leads us to replace relationships that are themselves part of what makes learning, healing, and social belonging possible.

What if we never learn the skills AI replaces?

Third, we need to think beyond “deskilling” to the danger of “never-skilling”.

If an adult uses AI to augment a capability they already possess, that may expand their agency. A student, however, may use AI and produce an essay but it is important that the task itself calls for human judgement before submitting. If we don’t add these steps, what happens when reliance on AI occurs before the underlying capability has developed?

We may create a generation able to produce sophisticated outputs without ever acquiring some of the judgement and critical capacities that previously made those outputs possible.

The issue is therefore not simply whether AI improves performance. Human development calls for people to develop capabilities for themselves over their lifetime, and they’ll need to be adaptive.

We need to decide which capabilities future generations should continue to acquire even when machines can perform the task more efficiently. And we need to direct the activity to be human.

What happens to human imagination?

Finally, there is imagination.

Human imagination is not simply the ability to generate something novel. My research across psychology, social theory, political theory and anthropology shows that imagination develops socially through play, language, and culture. It is intersubjective – something we do that is embedded in social practices and not just our mental cognition.

It allows us to see the world from another person’s perspective and, crucially, to imagine social arrangements that do not yet exist.

If AI increasingly mediates childhood, education, culture and everyday conversation, it will enter the very relationships through which imagination develops.

The danger is twofold. Firstly, that the human-to-human basis of our imagination shifts to a non-human AI partner. Secondly, that systems trained only on information that already exists may limit what humans believe is possible to change.

Why the UN matters

This is why the UN’s inquiry matters.

The central challenge is not to decide whether AI is simply “good” or “bad”. Nor is it realistic to imagine that the UN can control its development.

The more important role is to ensure that the future of AI is not decided by technological capability, corporate tech giants, and geopolitical power alone.

It should be that global coordination, genuine participation by smaller and developing states, protection of different knowledge traditions, investment in human capabilities, and a much broader conversation about what we want technology to augment – and what we want to preserve.

I argue that the UN Residence Coordinator system – which is the body that coordinates UN agencies’ country-level work under one senior representative – can play a leading role in this process.

Rather than governance, which is unrealistic in the tense geopolitical context, this system can help ensure there is coordination of AI developmental needs across all states.

The question I took away from Xi’an is therefore a simple one:

How do we ensure that increasing artificial capability enlarges, rather than diminishes, humanity’s capacity to understand, recognise and imagine one another – and collectively determine our own futures?
 
+ Dr Brincat was the sole Australian academic participant invited to this recent UN meeting on AI and Human Development