Extending a Causal Analysis Suite for Health Analyses: Capturing and Validating Critical Assumptions
Emre Kiciman
May 17, 2023, Wednesday, 3:00 PM - 4:00 PM EDT EDT
The role of a causal software platform is to scaffold the end-to-end causal analysis process to ensure that practitioners follow best practices in modeling and validation. In our experiences, based on usage of our causal tools (DoWhy, EconML, and other tools) across across industrial, agricultural, advertising, health and other domains, we find that data scientists and other practitioners are most challenged by the causal framing of their problems and validation of results. That is, errors in causal analysis are not (just) algorithmic, but in basic assumptions about a particular problem. In extending a causal analysis platform for health analyses, there is an opportunity to embed additional domain knowledge to better bridge the gap between health and causal tasks. In this presentation, I discuss what this can look like and our experiences in causal health analyses.
Emre Kiciman is a Senior Principal Researcher at Microsoft Research, where his research interests span causal inference, machine learning, and AI's implications for people and society. Emre is a co-founder of the DoWhy library for causal machine learning. He received his PhD in Computer Science from Stanford University.