Equal Opportunities

“Different perspectives are essential for AI”

The initiative Diversity for AI in Medicine teaches how to build fair and equitable AI applications that are helpful for humanity. The team behind the initiative has now been awarded the Prix Lux equal opportunities prize by the University of Bern.

Stavroula Mougiakakou: We started DAIM in 2022, while ramping up the Center for Artificial Intelligence in Medicine (CAIM), which virtually regroups all the researchers in Bern active in this field. The initiative was launched in response to global concerns about biases and inequalities in AI applications. We want to foster inclusion and diversity and believe that different perspectives are essential for driving innovation in AI, especially in areas as impactful as healthcare and medicine. We also want to encourage female researchers and researchers from minority groups to pursue a career in AI for medicine.

Monika Kugemann: The initiative is essentially conducted by two bodies: the DAIM committee and the DAIM community. The Committee is composed of three senior AI researchers, as well as an expert in ethics and myself, a communications specialist. We meet to plan and to arrange for the different activities, such as lunch talks or our big annual event in March. We also discuss outreach ideas to raise awareness, including joint events with the CAIM Ethics Lab.

The DAIM community is essentially open to everyone active in the field of AI in medicine with an interest in topics like inclusion, fairness or explainability or transparency in AI algorithms. We deliberately keep the focus wider than just «women in AI». We also partner with other initiatives of the Uni Bern, such as KILOF, the Bern Data Science Initiative, «Womxn who start up» and sitem-insel’s AI Symposium. Whoever wants to join, just needs to subscribe. The community is growing steadily and now encompasses more than 150 persons. We really want to thank the community for supporting us! Their active participation, their feedback and ideas are an invaluable help all along the way.

“We launched DAIM in response to global concerns about biases and inequalities in AI applications.”

Stavroula Mougiakakou

Mougiakakou: It’s not intangible. On the contrary, there is a plethora of examples of AI models that have biases and amplify stereotypes. Such biases are not built by purpose, they often are simply the result of ignorance or lack of attention. With a more diverse pool of people developing algorithms, we increase our chances to reach fair and unbiased solutions. Kugemann: Sometimes biases in the models arise, because the data on which the AI models are trained is one-sided or skewed: It represents only a part of our society. For example, women during their reproductive years have been excluded from many clinical studies for safety reasons. Thus, the results of these studies apply to male bodies. This has practical consequences. At an event about bias in AI and in medical research that we organized, an oncologist told us that immunotherapies work better in male than in female patients.

Kugemann: Yes, but being aware of gaps in the data can be a first step to remedy the situation.

Mougiakakou: Yes, on one hand, if you can’t avoid biases, you need to communicate them. You have to clearly mention what data has been used and what your model can and cannot do. It is unfair to generate wrong expectations. On the other hand, creating new AI models is a very active research field. We spend a lot of time and energy in developing algorithmic approaches that are able to detect, quantify and correct potential biases in the data they are being fed for training. We are also trying to develop methods that boost interpretability and transparency. We want to create AI models that are able to explain how they work and why they take a certain decision instead of another.

“Women during their reproductive years were excluded from many clinical studies. Thus, the results of these studies apply to male bodies. This has practical consequences.”

Monika Kugemann

Mougiakakou: There is a dedicated DAIM-award, which is given to research projects that provide innovative solutions to reduce biases and at the same time promote inclusive research practices. The awarded projects are a sort of best case examples showing that our aims are not just theoretical, but that they can be translated and put into practice. It is very important to us to foster and support interdisciplinary cooperation, because successful implementation of AI solutions in healthcare crucially depends on teamwork: The developers of algorithms, the physicians and the patients are like different pieces of a puzzle, and all these pieces need to be put in the right place to solve the puzzle.

Mougiakakou: At the end of the day, challenging issues related to discrimination is a matter of culture and of education. We need to create an educational system, in which students learn that all humans have equal rights. Over time, these values will be implemented in the AI solutions that they will develop. Therefore, we created our Master Program «Artificial Intelligence in Medicine», in which the students from the very beginning are concerned with the question of how to build fair and equitable AI applications that are helpful for humanity. The program started in 2021 and, in my eyes, it is a big success, not only because it attracts many students, also from abroad. But also because more than 50 percent of our students are women. This percentage far exceeds the global average, where women make up only 26 percent of the workforce in the field of AI.

“At the end of the day, challenging issues related to discrimination is a matter of culture and of education.”

Stavroula Mougiakakou

Kugemann: We are not entirely sure about the reasons for this extraordinary interest of young women in our Master Program. It likely has to do with the fact that the program offers a unique approach, providing students not only with a strong foundation in AI but also the opportunity to collaborate directly with physicians in the hospital. The master students, which all have a background in either engineering, computer science, mathematics or physics, gain important insights into everyday life on various hospital wards during their rotations at the Inselspital.

Mougiakakou: Maybe the interest is also linked to the fact that we have three female professors teaching in the Master Program. We may be role models and counteract the general perception that computers and algorithms are the domain of men. I don’t know if there is a secret ingredient. But in any case, our courses are organized and taught with a lot of care and enthusiasm.

Kugemann: The goal applies to both. There are AI applications that assist and empower patients to self-monitor their chronic disease, such as diabetes or different eye conditions. By helping them to better live with the condition, such AI solutions contribute to building equal opportunities for patients. Other AI applications support good and objective clinical decisions. Even at the end of very long days, AI systems don’t get tired or less attentive – and thus allow for a fair and equitable advice for each patient. However, AI is always a tool for the expert and does not replace him or her.

Kugemann: We are really happy to receive this prize! It is a very nice recognition for the hard work that we – and everyone else in the DAIM community – has put into this endeavor. We will use the money to enhance our portfolio in the next years. For what exactly stays a surprise for now, as the discussions are still ongoing.

Subscribe to the uniAKTUELL newsletter

Discover stories about the research at the University of Bern and the people behind it.

Subscribe to the newsletter