Elsevier Medical Education presents an interview with Gladys Lopez, Elsevier’s Global R&E ERG Chairperson and EMPOWER Founder and Leader (Diversity & Inclusion in Healthcare), on how AI can shape a more inclusive and patient-centered approach to training.
From your perspective, how can AI contribute to training healthcare professionals who are better prepared to care for diverse and inclusive patient populations?
Firstly, AI can identify and mitigate bias in healthcare training by insuring diverse and inclusive content is taken into consideration. This is essential for creating an equitable AI system that works across different social cultural environments. Then, the second part is that it can also increase their emotional intelligence. And you can ask, “How can AI create more empathy?” and the answer is that, by virtual simulations and increasing exposure to different characteristics, our students can train themselves to think outside the box and be exposed to different people that maybe, in daily life, they are not exposed to.
In what ways can AI-driven content ensure more equitable representation of different patient profiles?
AI could help save time here. A researcher can only analyze a certain amount of information per day, and the time and scope can be limited. With AI-driven content, good prompting and trusted content, a person can ensure that the references used are not only including a certain part of the population but also include a wider and more meaningful representation.
"AI can identify and mitigate bias in healthcare training by insuring diverse and inclusive content is taken into consideration. This is essential for creating an equitable AI system that works across different social cultural environments."
What risks exist regarding bias in AI data and algorithms, and how might these impact the training of future healthcare professionals?
We can take the example of female health here. Back in the day, the available bodies to understand Western medicine were white and male. Therefore, science advanced by taking these two characteristics as the norm. Nowadays, projects like MESSAGE in the UK and Europe are working towards adding sex and gender as a biological variable for research and policy. They are highlighting why these role models need to expand and include other intersectional areas. We currently have thousands of pieces of evidence for more diverse populations and this is exactly when AI can be risky if you don’t take trusted content as the base to fill AI.
What role do content providers, such as Elsevier, play in ensuring that AI systems used in medical education are inclusive, diverse, and free from meaningful bias?
Content providers like Elsevier have a huge responsibility for our society right now. Generative AI is a tool like a GPS in a car. If you give it the instructions, you trust that you’re going to be taken from point A to point B. However, you need to take into consideration two things. The first one is whether the GPS that you are trusting has enough information and knowledge about the surrounding territory. And the second is whether you are giving clear instructions. The same goes for generative AI. First, you need to trust that the content that is going into the AI tool is trustworthy. In that sense, Elsevier already has this trustworthy content and that is the reason a lot of people trust our tools. Also, we’re developing with inclusive content. The second part is that you are actually giving the AI tool sufficient information. In this case, it’s highly important that you learn how to prompt and that you are including different intersectional topics in your prompt to make it more inclusive.
How can AI help raise awareness among students about health disparities and support a more patient-centered and inclusive approach to care?
If you want to be more patient-centered and have an inclusive approach to care, you should ask yourself three questions every time you’re interacting with a generative AI tool. The first one is, “Is this content trustworthy?” The second one, “Is this answer representing me?” And the third one, “Is the answer accurate?” Then, I invite you to change the prompt including different intersectional topics, and you’ll realize how the content behind that answer can change and also, how the answer is adapting to all populations.
What competencies should students develop to identify and manage potential biases in AI tools during their clinical practice?
I would say communication and critical thinking training are key here. Regarding communication, the better the question, the better the answer. When it comes to AI, the better the prompting – including intersectional areas like sex, gender, ethnicity, disability status, among others – the better and more inclusive the answer will be. And the second one is critical thinking, and that comes after receiving the answer by knowing that it is only trustworthy if it’s backed up in evidence.
Looking ahead, how do you envision AI shaping the training of healthcare professionals capable of delivering truly inclusive and personalized care?
I envision a world where we can advance progress together. This world where students can easily identify information from misinformation. Where students are prepared enough to assist patients in a holistic way, including their diverse characteristics and connecting with evidence through generative AI that is based on trusted content. In this way, you will get the best diagnosis and treatment. This world, where AI serves as an expert advisor, along with the critical thinking of the future doctor, will help ensure quality and increase not only the life span, but also the health span for all.
Leave a Reply