We’re so focused on celebrating the loudest achievements in AI that we seldom ask ourselves: what are the intentions behind these solutions? We assume that every breakthrough in artificial intelligence is meant to advance our species, either for our productivity, economic growth, or efficiency. While all these benefits are commendable, we shall not forget that our future requires more than that: decision-makers who aim to prioritize responsible and equitable use.
Written by Viktória Vargová
Who is present at the meeting when a new algorithm is designed? Who is defining what “normal” means in a predictive policing tool? Why does a healthcare chatbot decide some symptoms are urgent while others are being dismissed?
Answers to such questions – and the thought-processes behind them – shape everything else that follows. They shape how LLMs interact with different backgrounds and how inclusive they become. You can rarely get true commitment to diversity if the ones calling the shots become painfully homogeneous.
AI is not a neutral technology. It reflects the assumptions, priorities, and blind spots of the people who create it. When those people are drawn from a pool that is all but varied (and the data shows they overwhelmingly are), the systems they build carry those limitations into every aspect they influence. This is not a hypothesis; it’s our empirical reality. And whether we like it or not, the consequences disproportionately affect those who had the least say in how these systems were designed in the first place.
Women represent just 22% of the global AI workforce. Fewer than 15% hold senior leadership roles. In the 111 countries that have Chief Information Officers, only 11% are women.1 When it comes to AI ethics research, one of the most consequential branches of the field, much of the groundbreaking work has come from women-led programmes and institutions. Paradoxically, those voices remain structurally underrepresented in the rooms where decisions are made.2
We all know this is not an inequality of opportunity. But it will present an inequality of outcome for all of us.
When computer scientist Joy Buolamwini discovered that the facial recognition software she was developing at MIT failed to detect her dark skin, responding only when she wore a white mask, she termed this finding the “coded gaze.” The datasets powering these systems were so skewed toward lighter-skinned, male subjects that they had become, in effect, a technical expression of existing social bias. The consequences of such systems are not theoretical: the technology replicated the very injustices it was supposed to be agnostic about.3
It’s a hard pill to swallow, but this is what happens when the design process lacks systematic diversity.
None of this means we should abandon AI as a force for change. The potential is real and significant; we just have to learn how to design systems that do not propagate an existing bias. For professionals who train their models on a wide array of samples and try to address bias before it occurs, the outcomes already yielded significant results. Take, for instance, solutions where AI-driven diagnostics are reaching underserved communities that traditional healthcare has failed. Or LLMs that are helping scientists model climate tipping points and optimize clean energy systems at a speed and scale that would be impossible without them. Or perhaps predictive tools that are enabling earlier intervention in public health crises. The technology can be powerful in many ways humankind hasn’t witnessed before.
But power without representation is precisely the problem. As researchers Ilcic, Fuentes and Lawler argue, AI systems exhibit characteristics of complex adaptive systems — non-linear, emergent, and deeply resistant to governance frameworks that assume predictability and control.4 In simpler terms, we cannot simply bolt ethical oversight onto a system built from the wrong foundations. The values embedded at the design stage propagate through everything downstream. Governance, as they frame it, must treat AI not as a technical artifact to be regulated afterwards, but as a socio-technical system that both shapes and is shaped by the social structures around it.5
Such a paradigm gives us implications for the levels of intersectionality that ought to be involved in building these systems from the start.
The ongoing narrative forces us to understand the twofold reality: AI is simultaneously our most powerful tool for addressing systemic challenges and one of the most potent amplifiers of systemic bias we have ever created. Addressing this dichotomy requires not just better algorithms, but more inclusive institutions and meaningfully different people in decision-making roles.
UN Women has been unequivocal on this point. The current gender digital divide, they argue, is “the new face of gender inequality.” The structural exclusion of women from digital governance – especially women from marginalized communities – means that the systems being built systematically fail to account for their needs and their lived realities. The result is technology designed for everyone but actually serving some far better than others.6
Awareness campaigns or diversity pledges might help, to some extent. However, tackling this systemic issue requires what sustainability professionals call participatory design: involving affected communities in how technologies are shaped and deployed, from conception through to accountability.7 Such an approach requires an additional layer of adaptive governance: making institutional frameworks flexible enough to evolve as the technology evolves, and inclusive enough to incorporate perspectives that dominant groups consistently overlook.8
AI equity requires collaboration on all fronts: between developers, researchers, and global policymakers. Only then can we work towards dismantling internalized prejudice embedded in discriminatory feedback loops, flawed algorithmic infrastructure, and bias in training data.9
There is something about initiatives putting women in AI into the spotlight that policy papers and governance frameworks cannot do: they recognize real work that challenges the status quo and enforces intersectionality from the start. They show the next generation of women in tech that the field has space for them, right at the frontier.
This matters more than it might seem. Role models are necessary to show people from underrepresented groups that their voice matters. When a young woman studying data science or environmental modeling or biomedical research sees someone who looks like her being recognized for work that changes things, it changes what she believes is possible for herself. Ideally, that shift in belief encourages more women to enter the field, stay in it, and eventually shape it. The visibility then becomes a part of the infrastructure of change.
The women doing this work are already out there. They are applying AI to predict biodiversity collapse or building diagnostic tools that catch diseases earlier than any clinician could alone. They’re developing language models that serve communities technology has consistently left behind, or using machine learning to make heavy industry greener and more sustainable.
Many of these women aren’t waiting to be discovered. They’re busy doing the work that matters. But we believe they deserve to be recognized publicly by institutions that take their work seriously.
The window for shaping what AI becomes is not perpetually open. Decisions being made today as new companies and products form will have consequences that outlast any individual policy or mere product cycle. The patterns being embedded now, in training data, in design assumptions, in the composition of the teams building these tools, will be extraordinarily difficult to reverse once they harden into a new level of infrastructure.
This is why the question of who is in the room is not a secondary concern to be addressed once the important technical work is done. It is the very urgent issue we need to address now.
Recognizing women who are already leading this change is a signal to our mankind that the future of AI is not a closed room. On the contrary, it gives a green light to expertise and authority that look many different ways. And showcasing that the technology shaping all of our lives should be built by people who reflect all of our lives.
Then make a difference and nominate her for the AI4Her Awards 2027. We recognize women reshaping the AI field across four categories: Health, Earth, Future and Legacy, and we are looking for women in the Benelux who are building AI responsibly, solving real problems, and leading with integrity.
Nominating takes three minutes and is open to everyone. The person you nominate will not put herself forward. But you can. And in that way, you’re changing the trajectory of how our future is being built. 👉 www.aifourall.org/awards/