Speaker: Tim Chen

Title: Beyond Ground Truth: Human-Centred Machine Learning for Noisy and Subjective Labels in Medical AI

When/where: Thursday November 27th, 2025 – 11h00 – Room 103

Abstract: Traditional notions of ground truth often fail in medical AI, where labels are inherently noisy and subjective. This talk reframes such “imperfections” as valuable human signals rather than errors to be removed. I will introduce human-centred machine learning frameworks that learn from this subjectivity, anticipate the needs of unseen users, and adaptively request expert input when uncertainty arises. Drawing on methods from Contrastive Learning and Learning-to-Defer, I will demonstrate how we can build AI collaborators that respect clinical expertise—moving beyond accuracy metrics toward more effective and trustworthy human–AI partnerships.

Bio: Dr. Tim Chen leads the Human-Computer Interaction Lab at the University of Adelaide, specialising in Human-AI Interaction and Spatial Computing. His research passion lies in engineering novel interactive systems using emerging technologies to augment the capabilities of knowledge workers. Dr. Chen is the human-AI interaction lead of the Eureka Prize-winning, MRFF-funded IMAGENDO project. He also serve as the University of Adelaide Research Lead for the Augmenting Ability CRC which aims to advancing adaptive AI solutions for aged and disability care. You can learn more about his work at https://ht-timchen.github.io/.