In the words of Oh Nam Kwon: "Justice must be designed into the code."

Oh Nam Kwon is the 22nd President of the Korea Federation of Science & Technology Societies (KOFST) and a Professor of Mathematics Education at Seoul National University. She also serves as President of the International Group for the Psychology of Mathematics Education (PME).

On 11 March, she spoke at the UN Women Knowledge and Partnerships Centre's International Women's Day commemoration in Seoul, Republic of Korea, reflecting on how algorithmic bias perpetuates gender discrimination and why women's participation in science and technology is a matter of justice, not opportunity.

"AI learns from data. That data is produced by our society. If society contains discrimination, so does the data, and AI learns that discrimination, faster and more precisely than anyone."

Kwon began by discussing a case from the United States. In 2013, a judge in Wisconsin used an AI system known as COMPAS to help determine a defendant's sentence. The system predicted recidivism at nearly twice the rate for Black defendants compared to white defendants. Historical and existing inequalities had been encoded into the algorithm.

This is not an isolated case. In 2018, Amazon scrapped its AI recruitment tool after finding that it consistently ranked women applicants lower. Trained on a decade of hiring data that largely reflected male candidates, the system had learned to replicate those patterns.

More recent research suggests that such bias extends beyond hiring. A 2025 study published in Nature, conducted by researchers from UC Berkeley, Stanford, and Oxford, analyzed 1.4 million online images across nine AI systems. It found that AI systems consistently depicted women as younger than men. When asked to generate CVs, one leading model reduced the stated age of female applicants by an average of 1.6 years and shortened their career histories. When those CVs were evaluated, older male candidates were rated more favourably, revealing how bias can shape both representation and assessment.

"Technology is a mirror. It reflects the society that built it, and the perspectives of those who built it. The problem is that this mirror is currently only facing one direction."

Who, then, is building the technology?

Globally, women account for around only 22 per cent of AI researchers. In Korea, the disparity is even more pronounced. While women make up 54.9 per cent of university graduates, they represent only 12.4 per cent of leadership positions in science and technology. Korea ranks 37th out of 38 OECD countries on this measure. At the current pace, it would take 22 years to reach the OECD average.

"The gap between 54.9 and 12.4 is not just a number. It reflects the stories of women who left their labs, the dreams they set aside, and the questions our society never got the chance to hear."

Kwon traces this gap back to the classroom. Over more than 30 years of teaching, and through collaboration with researchers in over 70 countries through PME, she has observed a consistent pattern. Girls perform strongly in mathematics in primary school. By secondary school, however, many come to believe it is simply not for them. Not because of ability, but due to their environment, teacher expectations, and the quiet accumulation of social messages telling them otherwise.

This has implications beyond education. Mathematics is the language of AI. Algorithms are built on mathematical structures. Who learns this language ultimately shapes who designs the systems that influence society.

"We don't need faster technology. We need better questions. And to ask better questions, we need people with diverse lived experiences – especially those who have historically been excluded – at the design table."

Kwon concluded by reflecting on the International Women’s Day 2026 theme: Rights, Justice, Action. In the context of science and technology, she described these as interconnected commitments. Rights mean that every woman and girl can learn and use the language of mathematics and science. Justice means that technology is built to incorporate diverse perspectives and produce fair outcomes. Action means starting now – in classrooms, research labs, policy spaces, and events such as this one.