Interactive lesson~18 minIntermediate
Causal Inference
Causal inference asks what would happen under intervention. It separates prediction from explanation by modeling how the world generates data.
DAGsDo-calculusCounterfactuals
Mental model
Correlation watches; causality asks “what if we changed it?”
Product decisions, policy, medicine, and model interventions require causal claims, not just predictive correlations.
Identification
balanced63% modeled signal
Bias control
balanced63% modeled signal
Actionability
balanced56% modeled signal
Concept pipeline
Build the idea in four moves
Interactive lab
Estimate whether a product change caused retention improvement.
Graph
Draw variables and assumed causal arrows.
Focus lens
The part that usually clicks late
Confounding
A third variable can create misleading correlation.
Identification
63
Bias control
63
Actionability
56
Knowledge check
Why is correlation not enough for intervention decisions?
Next horizon
Where this topic is headed
Causal representation learning
Uplift modeling
Causal evals for agents