AI, Power, and the Politics of Knowledge: Why Representation Is More Than Data

Most of us don’t design algorithms or train large language models. Yet every day, through our stories, our languages, and our curiosity, we influence what gets amplified and what gets buried. In today’s AI landscape, what gets amplified is disproportionately shaped by the Global North. This imbalance is not simply technical. It is structural.

Globalization: Neocolonialism or Enrichment?

Globalization is often framed as a force for integration and growth. But its effects are uneven. While it can enrich, it frequently replaces local narratives with globally dominant ones.

From food and fashion to language and education, global systems often normalize Western standards. For instance, global media disproportionately center Western perspectives. Over time, this creates a subtle but powerful shift: people in the Global South become more familiar with Western stories and ways than their own. AI systems, trained on this skewed representation, then reinforce the cycle.

Globalization and Dominant Knowledge Systems

At the heart of this issue lies a deeper tension among different knowledge systems. Those that mostly follow Western systems tend to prioritize analysis, measurement, and standardization. These are the foundations of modern science, including those in the humanities, as well as in social and natural sciences. But, many societies in the Global South are rooted in oral traditions, communal reasoning, and holistic understanding. Key anthropological findings suggest that oral cultures often prioritize dynamic, embodied, and relational knowledge over the fixed and linear forms typically valued in Western traditions. For example, intuitive practices—like cooking without fixed measurements or telling an age based on historical events rather than an exact date in the calendar or relaying direction based on wider landscape rather than a specific address point—reflect holistic, adaptive, relational knowledge.


I am speaking of the majority of countries in the Global South, not all. And, I am not denying the existence of ancient civilizations with literary cultures. Further, I want to emphasize that these systems are not inferior. They are different, and most notably, richer and more holistic. 


The dominance of Western institutions in education and research further reinforces this imbalance. Universities, journals, and AI labs—primarily located in the Global North—set the standards for what counts as valid knowledge. This creates a feedback loop: knowledge that aligns with these standards is amplified, while other forms are marginalized. A way to capture alternative ways of knowing in these systems has, to my knowledge, not been attempted. Rather, the same is repeated, solidifying and reinforcing Western ways of knowing. Boaventura de Sousa Santos adds: "besides not knowing the other kinds of knowledge, orthopedic thinking (which he defined as defined, measured knowledge) refuses to acknowledge [the very existence of other forms of knowledge]" Epistemologies of the South: Justice Against Epistemicide.


Scholars in postcolonial studies have also long critiqued this dynamic. The issue is not that analytical knowledge is wrong—but that it is treated as universal, rather than one approach among many.

Bringing AI to the equation

AI exists not in isolation, but within this structure. Universities, journals, and AI labs—primarily located in the Global North—set the standards for what counts as valid knowledge. There has not been an attempt to translate knowledge systems in the Global South into datasets. The current system can only depict discrete, standardized inputs. As a result, AI systems do not capture holistic ways of knowing. Artificial intelligence systems are also trained on vast datasets drawn largely from digitized, written, and standardized sources, which are dominated by Western institutions.


In addition, English and Western languages dominate online content, which directly affects AI training data. As a result, entire cultures are underrepresented or completely absent.

AI built on dominant power structures

This is not just a linguistic gap. Colonial legacies, patriarchal structures, and corporate power add salt to the wound. Global inequalities—visible in wars, economic policies, and technological access—are interconnected. 


What is often theorized in abstract terms is experienced acutely on the ground. It reflects deeper patterns of extraction. In the modern global economy, data has become a resource, often extracted from the Global South, processed in the Global North, then repackaged into products and systems that reinforce existing AI hierarchies.


The United Nations has called this a “data divide,” noting that countries in the Global South often lack control over how their data is collected and used. This gap between theory and lived reality is crucial. AI becomes detached from those whose data is exploited.

Access to advanced AI tools is also still uneven. Many platforms are unavailable or limited in parts of the Global South. One example is Claude, widely known as the best AI tool, which is unavailable in Ethiopia.

 A 2023 report by UNESCO highlights these concerns. The report emphasizes that without intentional inclusion, AI risks reinforcing existing inequalities rather than reducing them. 

As debates about AI ethics grow—focusing on representation, access, and exploitation—majority of voices, which are located in the Global South, remain excluded.

So Where Do We Stand?

The question is not simply whether the Global South should participate in AI conversations as the reality continues to be that AI will continue to evolve regardless.


The deeper question is: how can participation happen without erasure?


While structural change is critical and urgent, it often begins with something simpler: Telling our own stories. Valuing our own languages. Recognizing multiple forms of knowledge and honoring our own. Balancing holistic, relational understanding with analytical precision.