Cultural Intelligence in AI: Why Global South Thinking Matters

Artificial intelligence is often described as a universal technology, but that idea collapses under even light scrutiny. AI today is not global intelligence; it is Western cognition, scaled. Most debates focus on dataset balance or fairness metrics, yet the deeper issue is that worldviews themselves act as training data. A system built from a single cultural lens cannot hope to understand a plural world. Until AI learns from the cognitive architectures of the Global South, it will keep mistaking its own worldview for a universal one. AI Models must integrate all forms of human thought to fufill their potential.

Whilst I am not an ‘AI-guy’, I have spent the past five or six years working closely with artificial intelligence and am particularly interested in how its rise will reshape cultural and social biases

I argue often that one of the primary issues of western civilization has been that it thinks of its way of functioning as somehow ‘right' and other cultures must think in the same way to also be 'right'. Which is an absurd notion, but it is for many an unchallenged truth.

When thinking about these things I often notice the same pattern - we speak about artificial intelligence as if it were a universal system — a neutral machinery learning the world as it is. But AI today is not really learning the world. It is learning a world: the one contained in its training data, its designers, its assumptions, its cultural limits.

And today, largely that world is extraordinarily narrow in a global sense.

Almost every major model is built by young, highly educated, mostly white, men in the United States. Brilliant no doubt— but people carrying a specific worldview, shaped by a specific culture, with a specific relationship to identity, trust, hierarchy, risk, communication and indeed right and wrong - This is not a criticism. It is simply a fact. And because of that fact, AI is already biased long before it ever reaches the Global South.

And there are similar models being built in China, which have their own version of the same biases. My point however is not one of being anti this-or-that version, but more so one of being pro true global and cultural assimilation.

"Worldviews are training data. AI built on one worldview cannot interpret a plural world."

We know these Global South facing biases exist from simpler systems in banking. If an RM cannot pronounce a client's surname, they are passively considered a riskier proposition. If your country carries a broad prejudice, an individual faces personal barriers that have nothing to do with their own capability. These biases sit inside the seams of Western systems — employment, finance, governance and everyday interactions and they have become so ingrained as to rarely be intentional, but something worse, they have become structural.

Now imagine those same biases embedded inside systems that will make decisions for billions of people at a planetary scale.

This is why much of my work with Tintra focuses on cultural infrastructure — not just as a slogan, but as a way to move beyond the Western idea of “right” and “wrong” and use technology to shape how societies perceive identity, trust and behaviour, narrowing the delta between these worldviews that might seem disparate, but are actually not.

If modern AI is trained almost entirely inside a Western cognitive framework, then everything from fraud detection to credit scoring to identity verification, education systems, transport planning, healthcare diagnostics will inherit these world views.

And the people who will be misread first — and most often — are those furthest from that worldview.

So, to me, the Global South is not peripheral to the development of AI. It is the majority of humanity, only around 10% of the world live in ‘the west’ – so models being designed by that minority has obvious and inherent limitations. Non-western cultures, languages, patterns of thought, informal economies, communal structures and behavioural norms describe how the world actually operates. Yet these realities barely exist in the training data of most systems now being deployed across finance, governance, health, mobility and commerce.

This is the central paradox:

AI is not yet another technology the west sends to the Global South packaged as yet another saviour. It needs to be built with these cultures and worldviews from core or all we are doing is institutionalising the same issues even deeper than they already are.

AI today is largely an extension of Western cognition — linear, individualistic, Socratic in base logic – logic itself being Socratic of course. But human consciousness is not uniform. A Confucius minded China is a state that thinks through harmony, relationality and collective continuity. Africa thinks through an Ubuntu lens — personhood through community — a worldview that encodes identity, duty and trust differently. The Middle East thinks through lineage, honour and a form of negotiation. These are not surface differences; they are the operating systems of human intelligence, the Cultural OS we talk about in Tintra’s work.

So, for AI to be more than “I” — more than a machine reflecting a single worldview back at us — it must learn to operate across these different cognitive architectures. Until it can do that, it is not global intelligence; it’s just a narrow intelligence scaled globally.

It is easy to forget that human beings evolved inside small tribal units where knowledge was carried in relationships and shared memory. These systems worked well because they were built by humans for humans as we actually are, not as abstractions.

If human intelligence at its base form is a way for us to make sense of the present and predict the future, then the modern West has distorted that original version. Doing so through the belief that with enough laws, derivatives and optimisation tools we can replace or improve these evolutionary tools.

It is this same mindset that sits at the base of the modern western dissatisfaction – what amounts in many ways to essentially the individualised gamification of hope: you will be happy when you have the car, the house, the status, the likes – some predicted future that brings with it an abundance somehow missing in the present moment.

But real human intelligence has always been communal. Tribal societies manage risk not through documents but through trust, continuity and obligation to each other. And that’s not a primitive system; it is in fact a highly evolved one. And the gap between those two worlds — the tribal structures that shaped human evolution, and the modern systems that are so obsessed with prediction, is where it is essential that we don’t try to override the system of the many with the thinking of only the few.

By translating the mechanisms that kept those societies coherent historically into today’s infrastructure, we are very much able to build financial systems that reflect how humans actually behave, not how our models wish they behaved. That’s where the unlock sits.

When you work across the markets that we do you learn quickly that the constraint is never ingenuity. It is infrastructure — systems that were never designed with local realities in mind. AI risks repeating the same mistake. It asks billions of people to contort themselves into models that cannot see them as they are or contextualise their behaviour without distortion into a western framework.

At Tintra, we take a different view: systems should adapt to culture, not the other way around. As new technologies emerge, our task is to build AI-native financial and governance infrastructure that begins with anthropology, psychology and lived reality before it ever touches code. That includes indigenous knowledge — not as a nostalgic nicety, but because these foundational models of human interaction still shape how people communicate and cohere. The further we drift from them, the more fragmented our societies become - as we are seeing today across Europe and the US.

Whilst that might at first blush seem hard to reconcile, we should be careful to dismiss something that worked for a million years because we’ve acted differently for two hundred.

AI will not become global by scaling outward from Silicon Valley; it will become global by widening its lens. For the next decade of AI to fulfil its potential it must be shaped not only by technical progress but by cultural breakthroughs — by learning how to interpret collective decision-making, encode informal economies, read linguistic nuance where meaning sits between words rather than under a system, and  from there to build identity frameworks in societies where lineage, tribal affiliation or spiritual structure carry informational weight that Western systems cannot see - and are not designed to see. Moving away from obsession over one form of data at the expense of others

These are not edge cases I’m talking about here. They are the operating systems of most of the world

If AI is to serve humanity — and not just a subset of it — then the Global South must move from “market” to “model,” from recipient to reference point, from afterthought to centre of gravity. The world is reorganising into a multipolar landscape; technology must reorganise with it.

I close by positing that I believe that the future will not be defined by who builds the biggest model and then exports it, it will be defined by who builds the model that understands the world most completely in all of its diverse forms, cultures and ideologies.

Moving away from a world of one culture being right at the expense of another - to one where we use technology to close the gaps that we haven’t been able to do ourselves. Now that would be a noble use of and a genuine technological revolution – a system that allows us to understand each other better and brings us closer together rather than pushing us further apart.

About the Author

Dickie Shearer writes about cultural intelligence, emerging markets, and the evolution of AI in a multipolar world. His work explores how worldview, technology, and finance have the ability to shape the future. He is the founder of Tintra, a business focused on this work across the global south.