Causal Inference
Methods for determining cause-and-effect relationships from data, going beyond correlation to understand why things happen and predict intervention outcomes.
Key Concepts
Correlation ≠ causation: Ice cream sales and drowning both rise in summer, but one doesn't cause the other. Causal inference provides tools to distinguish genuine causal effects from confounding variables.
Methods
Randomized controlled trials: Gold standard but not always feasible. Instrumental variables: Using natural experiments. Difference-in-differences: Before/after comparisons. Do-calculus (Pearl): Formal framework for causal reasoning from observational data.
AI Applications
Understanding which features truly influence model decisions. A/B testing for recommendation systems. Medical AI (does treatment X cause outcome Y?). LLMs currently struggle with causal reasoning, an active research area.