FT online - tech and bias
AI medical tools
Here is another blog advising an easy way to find events in the world that are psychologically relevant and interesting, and link well with our new course. To use the psychology article cited in this blog, you need to visit FT online and sign up your students and yourself for free access to FT articles. Then you will be able to navigate to both the original FT article on AI technology and bias in health and well-being and my questions for students based on this.
FT articles
Original article from Sep 19th 2025 AI medical tools downplay symptoms in women and ethnic minorities
My article for a school DP psychology class related this original article to the health and well-being context, with a focus on the HL extension: technology and health and well-being and the concept of bias. Students should also be able to make links to responsibility. Answering the questions shows how psychology relates to current events and debates.
Once your students have read the original article, take them to the psychology class article and the questions. You may like to put them into small groups, as some of them will provoke discussion. Here are some suggested responses to the questions.
Suggested answers
1. What are medical professionals and social workers using the AI tools for?
To generate transcripts of patient visits, highlight medically relevant details and to summarise case notes. Some are also used in diagnosis.
2. What are the advantages of using AI tools for these purposes?
It reduces doctors’ and social workers’ note-taking time and speeds up treatment. Microsoft revealed it had built an AI-powered medical tool it claims is four times more successful than human doctors at diagnosing complex ailments.
3. What are some of the factors that are resulting in the downplay of symptoms in women and ethnic minorities?
Messages from medical professionals (taken from patients’ descriptions of their symptoms) that contained typos, used informal language or contained uncertain phrasing were more likely to generate AI model advice against seeking further medical care. This happened even when the clinical content was the same as in more authoritatively-written messages.
4. The article states that “A series of recent studies have found that the uptake of AI models across the healthcare sector could lead to biased medical decisions...” Follow the article link to one of these studies and summarise its argument in a few sentences.
Gourabathina et al. (MIT, 2025); Rickman (LSE, 2025); Zack et al. (Lancet, 2024) are the relevant studies. Answers may vary slightly, but gender bias and cultural biases in AI tools were found in all of these studies.
5. How does having several studies from reputable university departments, like the Massachusetts Institute of Technology, increase the reliability of the claims made in the article? What type of reliability does it increase?
These studies were looking at similar AI technology used in medical settings and evaluating the data created through similar summaries of medical reports, and show a high level of agreement. This suggests considerable inter-researcher reliability.
6. What role does stereotyping play in the bias shown in AI tools?
Stereotypes about ethnicity and gender affected the diagnoses and the treatment recommendations.
7. How do researchers suggest that the bias in medical AI tools could be reduced?
By training AI models on diverse and more representative health data sets.
8. What responsibility do researchers have when training AI models on real patient data?
Health data needs to be anonymised and kept confidential, and researchers need to be mindful that AI can “hallucinate” – make up answers – and that could be harmful.
9. Summarise your opinion about the present and future use of AI technology by health professionals. Choose at least one quote or example from the article to support what you say.
Answers will vary and could lead to a class debate or small group discussion.
References
Gourabathina, A. et al. (June 23-25, 2025). The medium is the message: how non-clinical information shapes clinical decisions in LLMs. FAccT ’25, Athens, Greece https://doi.org/10.1145/3715275.3732121
Heikkilä, Melissa (19th September 2025). AI medical tools downplay symptoms in women and ethnic minorities. Financial Times online, Accessed 6th October 2025 at https://www.ft.com/content/128ee880-acdb-42fb-8bc0-ea9b71ca11a8
Rickman, Sam (2025). Evaluating gender bias in large language models in long-term care. BMC Medical Informatics and Decision Making, 25, 274 https://doi.org/10.1186/s12911-025-03118-0
Zack, T. et al. (2025). Assessing the potential of GPT-4 to perpetuate racial and gender biases in health care: a model evaluation study. The Lancet, 6, e12-e22 www.thelancet.com/digital-health