We asked the same question to a faculty member from each of our three departments—Computer Sciences, Statistics, and the Information School. Read their full responses below.
Question: How do you use interdisciplinary collaboration in your research? Why is interdisciplinary collaboration impactful for research?
Yong Jae Lee
Associate Professor in Computer Sciences
My research would not be possible without collaborating with people from other disciplines, specifically medicine.
My recent research has looked at how physicians, nurses, and other people in healthcare interact with electronic health records. Electronic health records are complex pieces of software which can be used to view patient information, order medications, communicate with patients, and a lot more. A lot changed at the start of the COVID-19 pandemic as temporary restrictions to in-person care caused patients to send more electronic messages to their physicians to ask for care or advice instead of scheduling an in-person appointment. Our team has been studying the long-term impacts of this shift to computer-mediated care.
The only reason I can do this work is because I partner with physicians here at UW-Madison. My own training is in a field known as human-computer interaction which has its roots in a variety of disciplines including psychology, anthropology, and computer science. I like the diversity of this intellectual foundation, but there is a lot I don’t know about medicine. Since I came to UW-Madison four years ago I have been partnering with two primary care physicians who have been wonderful collaborators. By combining their knowledge of medicine and access to clinics with my knowledge of how people collaborate through software, we have been able to advance our understanding of how the growth in computer-mediated care has impacted not only physicians but also other members of the care team such as medical assistants and nurses.
Some questions don’t fit neatly within a single discipline. When systems are complex—social, technical, institutional—solving problems means starting with how they work and what they’re trying to achieve, then working backward to understand which assumptions, data, and modeling approaches will hold up in practice. That process shapes how we build models that are useful in the real world.
In my work, that often means designing AI systems that reflect how decisions are made in messy, real-world settings. One example is figuring out why high-risk patients get missed by medical systems. It’s tempting to frame that as a prediction task, but the challenge is often deeper. Data reflect how care is documented, not how it’s delivered. Clinical decisions depend on time, incentives, and uncertainty. Modeling those systems in a useful way takes more than standard tools. It requires methods that can reason under constraints and adapt to context.
I see modeling as a way of asking questions: surfacing assumptions, exposing blind spots, and clarifying what matters in the system we’re trying to understand. That’s what draws me to work that crosses disciplines. It pushes me to confront the limits of my own tools.
I often think about George Box’s line that all models are wrong, but some are useful. In practice, usefulness depends on whether the model reflects the system it’s meant to support. That’s where new methods and collaboration matter most.