Helping people design a better world.

Dr. Jutta Treviranus: The Human Starburst - Why AI Keeps Failing People at the Edges

Hey folks, and welcome back to another episode of This is HCD. My name is Gerry Scullion, and I'm a human-centred service design practitioner based in the beautiful city of Dublin, Ireland. Today's episode is a little different, and I'm genuinely excited about it. I'm handing the mic over to my colleague and guest host, Hyun Song, who sat down with someone I've wanted on this show for a long time: Dr. Jutta Treviranus, founder and director of the Inclusive Design Research Centre at OCAD University in Canada, and one of the most influential voices in inclusive design anywhere in the world. Jutta has spent more than four decades challenging the idea that we should design for the average person, and this conversation could not be more timely. In this episode, Hyun and Jutta get into what inclusive design actually means once AI enters the picture. They talk about why the data we train these systems on keeps failing the people who need the most support, why simply adding "more data" about people with disabilities can make an AI system more confident and more wrong at the same time, and what it will actually take to build systems that work for all of us, not just the statistical middle. Before we jump in, here are three things to listen out for: First: the Human Starburst. Since 1979, Jutta has been informally collecting one question from almost everyone she meets: what do you need to thrive? Plotted out, the answers form a starburst. Around 80% of needs cluster tightly near the centre, while the remaining 20% scatter across the huge, jagged periphery, and that's where people with disabilities tend to land. Nearly every system we design, from products to health care to education, works reasonably well for the centre and breaks down completely at the edges. Second: more data doesn't fix a biased AI, it can make it more confidently wrong. Jutta tells the story of testing self-driving car models with footage of a friend who pushes her wheelchair backwards through intersections, a real but statistically rare behaviour. Every system decided it was safe to proceed. The fix everyone suggested was more data. When Jutta retested with additional, even oversampled data, the systems didn't get safer, they got more confident that people in wheelchairs move forward, and would have run her over with greater certainty. Third: real inclusion means full participation, not personas. Jutta argues that people with lived experience of disability need to be in the room to frame the problem in the first place, not consulted after the fact through empathy exercises or diagnostic categories that flatten how varied disability actually is. She's chaired the technical committee behind Canada's new accessible and equitable AI standard, and she makes the case, backed by economic modelling, that designing for the whole spectrum from day one actually costs less over time, not more.Right, let's get into it. This is Hyun Song's conversation with Dr. Jutta Treviranus. Enjoy.