Theory and Methods Challenges in Counterfactual Prediction - Karla Diaz-Ordaz

Explore the intersection of AI prediction algorithms and causal inference in this 24-minute workshop video from the Alan Turing Institute. Delve into the limitations of traditional prediction methods and discover how causal inference can enhance decision-making capabilities. Learn about counterfactual prediction and its applications in healthcare and fairness assessments. Examine methodological challenges and potential solutions presented by a team of twelve academics specializing in predictive modeling, machine learning, and causal inference. Gain insights into decision support tools for scenarios like the COVID-19 pandemic. Cover topics including decision-making tools, causal assumptions, positivity violations, overlap issues, and real-world examples such as pneumonia treatment decisions.

Intro
Exercise
Decisionmaking tools
Causal assumptions
Lack of positivity
Random positivity violations
Poor overlap
Linear vs quadratic
Pneumonia example
Two more challenges
Conclusion