AI decides who you will become

AI learns from data, but that data does not represent everyone equally. Large language models are trained overwhelmingly on English-language text. This means the ideas and ways of thinking belong mostly to one tradition. Everything is viewed through an Americanised lens. When used globally, AI quietly reshapes how people express themselves.

Kristy Claassen argues the problem runs deeper than skewed data. In Ubuntu philosophy (a tradition rooted in sub-Saharan Africa) you become a person through community, culture and language. Flatten these non-Western cultures and languages into one globally recognisable version, and you are losing the process of becoming human.

Ubuntu is often summarised as "I am because we are." It is a philosophical tradition from sub-Saharan Africa that understands human beings as fundamentally relational. You’re not an isolated individual but exist through connections to others. Claassen used Ubuntu as a lens to evaluate AI ethics.

Claassen travelled to South Africa to run focus groups with Ubuntu-practising communities and asked them what values they associated with AI. “Mainstream AI ethics tends to focus on preventing harm, but they kept raising things like creativity and diversity,” she explains. "Imagine that AI is getting rid of your identity," one participant said.

Claassen does not have the answer yet. What she does have is a framework that asks different questions. Instead of "Is this AI system fair?", Ubuntu asks: Does this technology help you become more fully human? Does it strengthen the relationships and communities through which you develop as a person? Does it protect the diversity of cultures and languages that make us distinctly ourselves? These are not easy questions. But they are the right ones to start asking.

And some are already building alternatives. SURF in the Netherlands is developing GPT-NL, a language model trained on Dutch culture for local control, while South Africa's Lelapa.AI builds models grounded in African languages and contexts. "This is what it looks like when you stop assuming one size fits all," says Claassen. "You build AI that understands the cultures it serves, instead of flattening them."

Dr Kristy Claassen is a researcher, lecturer and AI ethics specialist. She holds master's degrees in both philosophy and theology, with a focus on technology. Her PhD research was carried out at the University of Twente as part of the Ethics of Socially Disruptive Technologies (ESDiT) consortium, funded by the Dutch Research Council (NWO). She was supervised by Prof Dr Ciano Aydin (University of Twente), Prof Dr Ir Peter-Paul Verbeek (University of Amsterdam), Prof Dr. Wijnand IJsselsteijn (Eindhoven University) and Dr Julia Hermann (University of Twente).

Claassen is also one of the founders of EthicEdge, a consultancy dedicated to advancing ethical AI across the African continent. EthicEdge combines academic research with advisory work, grounding AI ethics in African histories, values and futures.