Tutorial on Causality

The tutorial will be held on .


Tutorial: Causality

Jonas Peters


In the field of causality we are interested in answering questions like how a system reacts under interventions (e.g. in gene knock-out experiments). These questions go beyond statistical dependencies and can therefore not be answered by standard regression or classification techniques. While humans are very efficient in learning causal relations between few random variables, we require automated procedures in situations where many and/or high-dimensional data are available.

In this tutorial you will learn about the interesting problem of causal inference and recent developments in the field. The tutorial does not require any prior knowledge about causality.

  • Part I: We introduce structural equation models and formalize interventional distributions. We define causal effects and show how to compute them if the causal structure is known.
  • Part II: We present three ideas that can be used to infer causal structure from data: (1) finding (conditional) independences in the data, (2) restricting structural equation models and (3) exploiting the fact that causal models remain invariant in different environments.
  • Part III: If time allows, we show how causal concepts can be applied in the field of machine learning.


Jonas Peters from Max Planck Institute for Intelligent Systems