WASP industrial PhD student Martin Lindvall is having his halftime seminar. A halftime seminar is the equivalent of a licentiate seminar, where Martin will be presenting his ongoing PhD work and discuss it with docent Ylva Fernaeus of KTH, our external discussant.
Designing for machine learning in medical image diagnostics — Enabling efficient ensembles of professional medical practitioners and learning machines through interaction design
Advancements in machine learning (ML) has led to a dramatic increase in AI capabilities for medical diagnostic tasks. However, despite impressive technical advances, predictive algorithms are to a very small extent used in healthcare today. A lack of design considerations and inefficient human-AI collaborations has been identified as a major factor limiting adoption in clinical settings.
Using a constructive design research approach, my work explores how we might design systems with ML components that aid clinical decision-making. Preliminary results are derived from experiments and experiences from four projects, all aiming to produce novel interactive systems for or with ML components.
Preliminary contributions consist of identifying design challenges and suitable interaction strategies for human-ML collaboration at different stages of automatic support. Specifically, I present three works that address challenges in designing for effective human-machine teaching and two in-progress works that address the challenge of designing interactions that afford successful collaborative strategies despite the uncertainty inherent in machine predictions.