Original title: Disaggregating the AI elite: Gender, age, and education differences among professionals engaged in AI development
Matti Nelimarkka · September 2026
Artificial intelligence shapes our world, so we need to know who builds it. Researchers asked who truly designs and controls these tools. They surveyed engineering and technology professionals to identify the AI elite, a small group inside the broader coding workforce. They tracked who worked on AI projects and analyzed what traits predicted that role. The study revealed that very few people hold real power over AI. Having a top job at work was the strongest factor, showing that people with existing power gain control over new tools. Gender kept women out both directly and by limiting access to education and senior roles. Surprisingly, using many different everyday AI tools actually linked to less involvement in building AI systems.
Imagine a group of friends planning a road trip. Everyone talks about possible destinations, but only the person who holds the car keys actually decides where you go. The others might suggest places, but they never get to steer. In this paper, the 'AI elite' are the ones holding the keys—only 5% of tech workers actually decide how AI is built—and the study shows that the keys tend to go to the same people who already had power: senior employees, men, and the highly educated.
Who builds AI determines what problems it solves and who it serves. This study reveals that only a tiny fraction of tech workers—just 5%—actually shape AI systems, and that group overwhelmingly consists of senior, educated men. That means the same power imbalances that exist in society are being baked into AI from the ground up, making diversification efforts urgent.
The study relies on self-reported survey data, which may not capture actual contributions. It does not establish causation—only correlations. The sample is limited to engineering and technology professionals, so findings may not generalize to other roles involved in AI. The surprising negative correlation between everyday AI tool use and development participation requires further explanation.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Disaggregating the AI elite: Gender, age, and education differences among professionals engaged in AI development — Matti Nelimarkka — SocArXiv — doi:10.31235/osf.io/d9e8k_v1
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper. This is a preprint, so it hasn't necessarily been peer reviewed.