The Bernoulli Institute AI Colloquium is a regular meeting where faculty and students can hear and discuss the current research related to the institute's three research themes (Computing and Cognition; Geometry and its Applications; Systems, Data and Society) from inside and outside the University of Groningen.
Colloquia are scheduled monthly during the teaching blocks, with the exception of December, preferentially on Thursday afternoon. Meetings at the end of the day are followed by the PoCoBo - a post-colloquium borrel.
For external colloquia, there is a limited opportunity to have dinner with the speaker in the evening. If you are interested in this, please contact Harmen or Stephen a few days before the colloquium.
The design of AI systems to assist human decision-making typically requires the availability of labels to train and evaluate supervised models. Frequently, however, these labels are unknown, and different ways of estimating them involve unverifiable assumptions or arbitrary choices. In this talk, I introduce the concept of “label indeterminacy” and derive important implications in high-stakes AI-assisted decision-making. I present findings from an empirical study in a healthcare context, focusing specifically on predicting the recovery of comatose patients after resuscitation from cardiac arrest. This study shows that label indeterminacy can result in models that perform similarly when evaluated on patients with known labels but vary drastically in their predictions for patients where labels are unknown. After demonstrating crucial ethical implications of label indeterminacy in this high-stakes context, I discuss takeaways for evaluation, reporting, and design, as well as potential ways forward.
Peter Hendrix, Tilburg University
Tsegaye Misikir Tashu, University of Groningen
Catherine Sibert, University of Groningen
Freek Stulp, Institute of Robotics and Mechatronics