A new feature in the Swiss daily Le Temps spotlights the work of Shin Koseki, UNESCO Chair holder in Urban Landscape at Université de Montréal and affiliate of Mila, exploring how artificial intelligence can help reveal who gets overlooked in city planning. Trained at EPFL, Koseki brings a perspective that resonates directly with the Swiss urbanism community the article is written for.
The piece describes Koseki’s approach to AI as a socio-technical system rather than a neutral technology, one that carries the assumptions and blind spots of the data and people behind it. It details two strands of his team’s work: participatory tools that score street-level imagery for accessibility and comfort using data generated with the people most affected by exclusion (older adults, people with disabilities, women, LGBTQ+ residents, immigrants, and others), and alignment techniques aimed at correcting generative AI models so their outputs better reflect marginalized perspectives, alongside non-verbal participatory methods that let residents shape urban imagery directly.
The article closes on Koseki’s argument for locally-grounded AI: models trained on North American urban imaginaries cannot capture the nuance of a street in Lausanne or a square in Geneva, and any Swiss adaptation of these tools should be methodological, not just technological, built with local communities, universities, and administrations.
Read the article: https://www.letemps.ch/societe/quand-l-ia-revele-les-oublies-de-la-ville