Elsevier Medical Education presents an interview with Gosia Warminska-Marczak, Director of Commercial Healthcare Education (DACH, Eastern Europe and France) at Elsevier, exploring the underrepresentation of women in AI, what’s needed for truly inclusive medical AI tools, and how medical education can embrace AI without losing the human element.
What role do women currently play in the development of AI technology and Health Education? Do you believe they are adequately represented?
What the data tells us about women’s role in AI
Women play a very important role, but I don’t think they are well represented – and the numbers show this clearly. According to the World Health Organization, women make up 70% of the global health workforce but when we look at who actually builds AI, the picture is very different.
According to data available on SheAI, women only hold around 22% of AI-related roles globally and less than 14% of senior AI leadership roles. And according to UNESCO, only about 12% of AI researchers globally are women.
Let’s look specifically at Europe now. According to the World Economic Forum, women’s share of tech jobs in Europe has dropped from 22% to 19% since 2023. And according to the International Labour Organization, roles predominantly held by women are almost twice as likely to be exposed to disruption from AI.
Women are adopting AI, but do they have equal power to shape it?
But there is also good news. According to Deloitte, US women’s adoption of generative AI has tripled in the past year – faster than men. So, women are engaging with AI a lot. The real question is whether we have equal power to shape it.
"If the data and design do not reflect women, the tools work less well for half the population. That is why we need women at the design stage, not only as end users."
Gosia Warminska-Marczak, Director of Commercial Healthcare Education, Elsevier | Elsevier Faculty Hub, July 2026
What can a female perspective bring to the digital transformation of healthcare? Are there still gaps or limitations when it comes to gender representation and perspective?
Why gender diversity in AI design is crucial in healthcare
Female perspective can bring quite a lot, not only in terms of fairness but also in the quality of the tools. Healthcare is deeply gendered. Diseases can present differently in women and men. Treatment responses differ and care priorities differ too. So, if the people building AI tools are not representative, these differences get “baked in” from the start and then repeated at scale.
Women often bring a more contextual, relationship-centered approach, and in a clinical setting, this matters a lot. There is also the lived experience of being a patient and navigating the health system, which shapes how you think about usability and access.
The gender gap in healthcare is documented – and significant
And the gaps are documented. A report on the health gender gap in Europe found that 22% of women in the European Union feel that men are treated better by medical staff – that’s nearly 1 in 4. And according to research on bias, some AI diagnostic tools have shown lower accuracy for Black women, partly because clinical trial data has historically been skewed towards men.
So, this is not theoretical. If the data and design do not reflect women, the tools work less well for half the population. That is why we need women at the design stage, not only as end users.
Where does the use and implementation of AI currently stand in medical education? Is it already a reality, or is there still significant progress to be made?
AI in medical education is already here, and student adoption has surged
It’s definitely already happening. This is not the future – it is the reality in the classroom now. Students use AI tools whether their institutions have formal policies or not. According to one US study, student adoption jumped from 24% to 77% between 2024 and 2025. Today, it is probably closer to 100% in different parts of the world.
The question for medical education is no longer if, but how to use AI
So, the tools are here. The big conversation now is not if, but how – how to use those tools responsibly and how to protect critical thinking. For me, one principle cannot change: a doctor is ultimately responsible for the diagnosis and the treatment. That accountability cannot be delegated to a machine. AI can support the doctor, but it cannot replace that responsibility.
Medical education therefore has to do three things at once:
- Embrace AI as a learning tool
- Reinforce clinical reasoning
- Recognize that the human element is more important than ever before
Formal AI teaching in medical schools is lagging far behind student use
But here is the problem. Most students still receive very little formal AI education. Most faculty recognize that AI is important, but few feel they understand it well enough to teach it. Students are using these tools daily, while formal teaching lags far behind. Closing that gap is one of the most urgent tasks for medical schools today.
What are the key factors that determine – or ensure – the implementation of reliable AI in medical education?
Reliable AI starts with trusted, peer-reviewed content
For me, it comes down to a few things. The first is the content behind the tool. AI has to be grounded in trusted, peer-reviewed content. This is where publishers and academic institutions have a real role because the quality of the output depends on the quality of what goes in.
AI hallucination in healthcare is more than just an inconvenience
The second is hallucination – this is one of the biggest risks. A system can generate something that sounds very confident but is simply wrong. In everyday life, this is annoying. In medical context, it can be truly dangerous.
AI literacy cannot be an optional add-on
The third one is AI literacy. I really believe AI literacy needs to become a core clinical skill, not just an add-on. Students need to understand how AI works, where it fails, how to question what it produces and to not accept the answer just because it looks professional.
What institutions must do: AI policies and quality assurance
On the institutional side, there are two more factors. First, clear policies defining where AI is appropriate or where it is not. Second, proper quality assurance before any AI is used in teaching. We would never introduce a medical device without testing it so we should treat AI in the exact same way.
Reliable AI in medical education is not down to any one factor. It is a combination of trusted content, awareness of the risks, skilled users, clear rules, and proper checks before deployment.
What would define an inclusive AI – one that truly accounts for diversity – in the context of medical education?
Inclusive AI requires training data that reflects all patients
It has to start at the very beginning with the training data. The data needs to reflect the full diversity of real patients – different sexes, ethnicities, ages, and geographies – because a tool can only learn from what it has seen.
You cannot fix a bias you don’t measure
There is one rule I always come back to: you cannot fix a bias that you don’t measure. Sex-disaggregated and race-disaggregated data needs to become standard, not an exception. Only then can we see where the problems are.
The next step is bias auditing. This should be a requirement before deployment – not something we check afterwards, once the AI tool is already being used to teach students.
Then there is the educational content itself. It needs to draw on diverse patient cases, like I mentioned before – different ages, ethnicities, geographies, and so on.
Diverse teams building AI tools are essential
Finally, the teams. We need diverse teams of people building these tools – women, clinicians from different contexts, educators from different healthcare systems – because the stakes are high. Biased AI in medical education does not just produce one unfair tool, it trains the next generation of doctors in biased patterns. If we get it wrong, we teach that mistake to thousands of future doctors. That is why inclusion is not optional.
"The challenges are significant, but I remain optimistic. If we involve the right people and keep the human element at the center, AI can truly improve how we train future doctors."
Gosia Warminska-Marczak, Director of Commercial Healthcare Education, Elsevier | Elsevier Faculty Hub, July 2026
Looking ahead, what are the main challenges facing the use of AI in training future healthcare professionals?
Students and educators need to understand the limitations of AI
I will group them into a few areas. The first is AI and digital literacy. Students and educators need to move beyond using AI as a better search engine and really develop an understanding of its limitations.
Over-reliance on AI risks undermining clinical reasoning
The second is protecting critical thinking. There is a real risk of over-reliance. If students use AI to shortcut the reasoning process, they skip the mental work that actually builds expertise. That reasoning takes time, but it is how you become a good doctor.
AI cannot teach empathy – and we must not deprioritise it
The third challenge is very important in my eyes: the human touch. Empathy, communication, cultural sensitivity, being truly present with a patient – AI cannot teach those things, and there is a risk that we deprioritize them simply because they are harder to measure.
The fourth is ethics, including data privacy, liability, and knowing where the boundaries lie. Many of these questions are still being worked out and educators need to navigate them carefully.
Access and equity: will AI deepen existing inequalities?
The last challenge is access and equity. AI tools cost money and require infrastructure, and not all institutions or countries have equal access. There is a danger that AI could deepen existing inequalities.
So, the challenges are significant, but I remain optimistic. If we involve the right people and keep the human element at the center, AI can truly improve how we train future doctors.
Leave a Reply