The Book of Why
Causal questions require assumptions about interventions and counterfactuals, not simply finding correlations in large datasets.
Association describes observed relationships without establishing their causes.
Two things moving together does not tell us what would happen if we changed one. Pearl makes that difference central, separating observations from the causal questions decisions actually require.
Intervention asks what changes when an action is imposed.
Counterfactuals ask what would have happened under an alternative condition.
These form distinct levels in the proposed causal ladder.
Diagrams express assumptions about causal relationships.
A causal diagram is useful because it exposes assumptions, not because drawing arrows proves them. Good reasoning asks which relationships are justified and what evidence could distinguish competing accounts.
Confounding can make observed associations misleading.
Some adjustments help while others introduce new bias.
Data alone cannot identify every causal structure.
Experiments and defensible assumptions supply additional information.
The conclusion makes causal reasoning an explicit part of explanation and decision.