Today’s blog post comes from Jenaya James, our intern, who recently completed her Master’s in Global Health at King’s College London in the UK. This is Jenaya’s second article in a series exploring topical public health issues.
In today’s blog, Jenaya takes a closer look at the growing role of AI in healthcare, exploring the opportunities it presents alongside important questions around innovation, ethics and trust.
Artificial intelligence, or AI, is becoming a bigger part of our everyday lives, especially in healthcare. It can help doctors read medical images and offer health tips through digital tools. AI has the potential to make healthcare better and easier to access for everyone.
The World Health Organization (WHO) states that AI can help with diagnosis, treatment, health research, medicine development, and public health. But it also warns that these benefits will only happen if ethics and human rights stay at the heart of how these technologies are made and used. This brings up an important question for healthcare: just because we can use AI, should we always use it?
What could AI offer healthcare?
AI is a term for computer systems that can do tasks that usually need human intelligence, like finding patterns, understanding information, creating language, or making predictions.
In healthcare, AI can be used in many ways. It can help doctors review medical images, spot patterns in large amounts of patient data, support decisions, and even handle paperwork. AI also has a role in public health and research. These uses are especially important for global health. Many countries, including those in the Caribbean, struggle with too few healthcare workers, high costs, and limited access to specialists. If used carefully, digital tools like AI could help health professionals do more with what they have and reach people who need care the most. However, new technology does not always make healthcare fair for everyone. AI systems depend on the data and choices made by their creators.
When an algorithm inherits our inequalities
A major concern with AI in healthcare is bias. AI learns from the data it receives. If the data does not represent everyone, the technology might not work equally well for all groups. This can lead to existing healthcare inequalities being built into new systems. Racial bias is a key concern in healthcare AI. If systems are trained on data that is not diverse, they may perform less accurately for some racial and ethnic groups. One U.S. study found that a healthcare algorithm underestimated the needs of Black patients because it used healthcare spending as a proxy for health needs. This highlights the importance of ensuring AI is tested on diverse populations, including those it is intended to serve.
This is important because health data are shaped by society. Differences in access to care, diagnosis, and treatment can affect the information in these datasets.
So the ethical question is not just whether an AI system is accurate overall. We also need to ask: accurate for whom?
If a system works very well for one group but poorly for another, it could makehealth inequalities worse. Developers and healthcare organisations need to check who is included in training and testing data, and see if results differ between groups.
The U.S. Food and Drug Administration (FDA), alongside international regulatory partners, has similarly emphasised transparency around training and testing data, known biases, limitations and populations that may be underrepresented when machine-learning medical devices are developed.
Privacy, consent and our health data
AI relies a lot on data, and few types of information are more personal than our health details. Medical records can include details about diagnoses, medications, mental health, reproductive health, and other sensitive parts of people’s lives. This raises important questions about privacy and consent.
Patients should know how their information is used and protected. However, AI development often involves large datasets and complex relationships among healthcare organisations, researchers, and technology companies. This can make genuine informed consent difficult. The rise of generative AI creates additional concerns. WHO has warned that users may provide sensitive health information to large language models without fully
understanding how that information may be processed or protected (WHO, 2023). Healthcare organisations need clear rules about what information can go into AI systems, how patient data are stored and protected, and who can access it.
Do we understand how AI makes decisions?
Imagine an AI system tells you that you are at considerable risk for a certain illness.
Your next question is: Why? In healthcare, this question matters enormously.
Some advanced AI systems can give answers without explaining them in a way that patients or even healthcare workers understand. This is often called the ‘black box’ problem. Transparency and explainability are important ethical principles. Healthcare professionals need enough information to decide when an AI recommendation is right, and patients should understand how technology affects important decisions about their care.
Transparency does not mean every patient must understand the math behind an algorithm. It means giving clear information about what a system does, what data it uses, how reliable it is, and what its limits are.
Who is responsible when AI gets it wrong?
Another tough question is about accountability. If an AI system suggests something that harms a patient, who is responsible? Is it the doctor who followed the advice, the hospital that bought the technology, the people who built it, or the company that sold it? We cannot allow responsibility to disappear behind an algorithm. WHO identifies protecting human autonomy and ensuring responsibility and accountability among its core principles for ethical AI in health. AI should therefore support rather than automatically replace appropriate human judgement, particularly where decisions can significantly affect a person’s health or wellbeing. This also means healthcare workers need to truly understand the technology they use. Relying too much on computer advice can be risky if people start to believe the computer is always correct.
AI can sound confident and still be wrong
Generative AI introduces another challenge: misinformation. Tools like large language models can give answers to health questions that sound convincing and professional. But just because something sounds right does not mean it is correct. WHO has warned that large language models can produce responses that appear plausible and authoritative while containing serious errors. Premature reliance on untested systems could therefore contribute to patient harm and undermine public trust. For the public, digital health literacy is becoming more important. People need to know that AI-generated health information should not be taken as medical advice just because it sounds confident.
The Caribbean and the global AI divide
Ethical AI is also a global-health issue. Most advanced AI technologies are developed in wealthier countries and by large tech companies. This raises questions about which populations, languages, healthcare systems, and priorities shape these technologies. For smaller states and Caribbean communities, using AI systems developed elsewhere without proper local testing could cause problems. A tool that works well in one population may not work the same way in another. At the same time, countries with fewer healthcare resources should not miss out on technology that could help them. The challenge is to avoid two extremes: using AI just because it is new, or rejecting helpful technology because of the risks. Instead, communities should have a real voice in how new health technologies are introduced.
Concerns of environmental costs
AI also has a harmful environmental impact. Training and running large AI systems demands a great deal of computing power, which in turn uses a lot of electricity and water for cooling data centres. As healthcare becomes increasingly reliant on AI, its potential advantages must be weighed against its environmental costs. For sustainable healthcare innovation, one should consider not only whether a technology improves care but also whether its resource use is justified by the benefits it provides.
Keeping people at the centre of healthcare
It is clear that AI will have a bigger role in healthcare in the future. The real challenge is to make sure that all this progress keeps people’s health and wellbeing as the top priority. WHO’s ethical framework emphasises six principles: protecting autonomy; promoting wellbeing, safety and the public interest; ensuring transparency and explainability; fostering accountability; ensuring inclusiveness and equity; and developing AI that is responsive and sustainable.
These principles show that innovation and ethics can go together. AI can help healthcare workers do their jobs better, spot things people might miss, and give more people access to good health information. But technology should support, not replace, the knowledge, care, and judgement that people bring. So perhaps the most important question about AI in healthcare is not whether computers will become smart enough to change medicine. It is whether we will be wise enough to make sure these changes help everyone.
References
- Obermeyer, Z., Powers, B., Vogeli, C. and Mullainathan, S. (2019) ‘Dissecting racial bias in an algorithm used to manage the health of populations’
- U.S. Food and Drug Administration (FDA) (2024) Transparency for Machine LearningEnabled Medical Devices: Guiding Principles.
- World Health Organisation (WHO) (2021) Ethics and governance of artificial intelligence for health: WHO guidance. Geneva: World Health Organisation.
- World Health Organisation (WHO) (2023) Regulatory considerations on artificial intelligence for health. Geneva: World Health Organisation.
- World Health Organisation (WHO) (2023) WHO calls for safe and ethical AI for health. 16 May 2023.
12
