Seminr: Representation learning for causal questions in science
By Francesco Locatello

Abstract

Deciphering experimental observations into structural knowledge of the world is a key component of the scientific discovery process and a longstanding challenge for AI. In this talk, I will present our recent works bridging between perception and causal knowledge. I will begin with the challenge of training predictors that are causally valid proxies of latent variables, from a theoretical perspective first and next in real world experiments in ecology. Next, I will also show how AI models enable “looking at the data first” and discovering effects in randomized trials without supervision.

Biography

Francesco Locatello is a tenure-track assistant professor at the Institute of Science and Technology Austria and is involved in the Google Research Project. Previously, he was a senior applied scientist at Amazon Web Services. He received his PhD from ETH Zürich, where he was co-advised by Gunnar Rätsch and Bernhard Schölkopf. His research has received several awards, including the ICML 2019 Best Paper Award, the Hector Foundation Award from the Heidelberg Academy of Sciences in 2023, and the Google Research Scholar Award in 2024.