Jonas Hanselle, M.Sc.

PhD student, Research assistant

Chair of Artificial Intelligence and Machine Learning

Office address:

Akademiestraße 7

Room 109

80799 Munich

Send an email

+49 (0)89 2180 73447

Research Focus

I am working on trustworthy decision support — the question of what it takes for a machine learning system to be reliably deployed in high-stakes settings such as medicine, policy, and science. For such support to be trustworthy, two things must hold: the model must have learned reliably from data that may be noisy, imprecise, or subject to distribution shift; and its outputs must be structured so that human decision-makers can understand, inspect, and act on them.

My research addresses both. On the interpretability side, I develop sparse, hand-computable models that are simultaneously interpretable and probabilistically calibrated, suitable as direct interfaces for human decision support. On the reliability side, I work on learning from imprecise, set-valued supervision, distribution-free uncertainty quantification via preference-based conformal prediction — using preference learning methods in place of hand-crafted nonconformity scores — and the unification of robust supervised learning methods against distributional shift.

Selected publications