Seminar • Artificial Intelligence • Discovering Knowledge via Causal Inference and Representation Learning

Please note: This seminar will take place in DC 1304 and virtually over Zoom.

Jason Hartford, Postdoctoral researcher
Mila, Quebec

Causal inference provides a powerful suite of tools through which economists, epidemiologists, and the social sciences understand the world. But textbook causal inference methods limit the questions that scientists can ask because they rely on classical statistical estimation techniques.

In the first part of this talk, I will discuss how we can make these tools more flexible by combining a classical causal inference approach—instrumental variable estimation—with deep network-based estimators, and how we can make them more robust by leveraging multiple instrumental variables. The second part of my talk will focus on a key question for the future of causal inference: how can we make causal inferences about high-dimensional unstructured data? These modalities, such as images or sensor data can often be collected cheaply in experiments, but they are challenging to use in a causal inference pipeline without extensive feature engineering or labelling to extract underlying latent factors. The goal of causal representation learning is to find appropriate assumptions and methods to disentangle latent variables and learn the causal mechanisms that explain a system’s behaviour. I will present results from a series of recent papers that describe how we can disentangle latent variables with identifiability guarantees. Finally, I will conclude by discussing my future research directions with a focus on the role of experimentation, and how we can go beyond the stringent assumptions that are currently needed to attain identifiability guarantees.

Bio: Jason Hartford is a postdoc with Prof Yoshua Bengio at Mila where he works on causal representation learning. Before joining Mila, he completed his Master’s and PhD at the University of British Columbia with Prof Kevin Leyton-Brown where he worked on deep learning-based estimators of causal effects, and enforcing symmetries in neural networks. During his PhD he had internships at Microsoft Research New England and Redmond where he worked on deep learning approaches for causal inference.

To attend this seminar in person, please go to DC 1304. You can also attend virtually using Zoom at https://uwaterloo.zoom.us/j/91039097830.

For those attending virtually: The passcode will be provided by email on Friday before the seminar as well as on the morning of the seminar.