Introduction to Causality

Speaker: Jakob Zeitler ( University College London, UK)

Title: Introduction to Causality

Abstract: Causal inference is of rising interest to many scientific fields where there is a need for explanations beyond correlation tables. As a well established method in econometrics and epidemiology, recent applications of it in machine learning have caught attention and sparked ideas on new approaches to fields such as healthcare and algorithmic fairness. This tutorial provides the necessary introduction to the paradigms of causality, both graphical models and the potential outcomes framework. The first session covers the key fundamentals of causal effect estimation and discovery. A further practical session facilitates hands-on experience and discussion of possible applications and current research to motivate further independent study. Ultimately, this tutorial aims to inspire the audience to pursue research in causal inference to develop interpretable and explainable machine learning, a very open area of research.

Bio: As part of my PhD at UCL, I research how causality can be used in machine learning and the other way around. Specifically, I am interested in sequential decision making and how to optimise it with ideas from causal inference. For example, in applications of reinforcement learning to optimisation of treatment regimes for hospital patients in intensive care units, causal relationships learned from historical treatment data can play a crucial role.