Judea Pearl clarifies: do-calculus presumes a causal model; learning it is causal discovery
yudapearl · x · 2026-09-19
- Answering a question about whether do-calculus "proves causation" (quoting a thread with Bernhard Schölkopf), Judea Pearl explains the field's deliberate split:
- Causal inference: starts from a (partially specified) causal model and answers causal questions given that model — do-calculus lives here, so it assumes rather than proves the causal structure.
- Causal discovery: going from finite samples to distributions (or distributional parameters), i.e., learning the causal model itself from data.
So the complaint that do-calculus doesn't prove causation misses the division of labor: recovering causal structure is causal discovery's job; do-calculus reasons about interventions once a structure is assumed.
More from Research
- USRA details NASA-IBM lunar foundation model fusing 11 lunar data modalities — victormustar · 2026-09-19
- NASA and IBM open-source lunar foundation model pretrained on 2M multimodal tiles — victormustar · 2026-09-19
- Stabilizing RL with a simple second-year probability trick: the covariance identity — PMinervini · 2026-09-19
- willcb: in-context learning for personal continual learning needs neurosymbolic harnesses — willcb · 2026-09-19
- Online RL via task flywheels from live interaction traces is what actually works today — willcb · 2026-09-19
- Tobi Lütke sparks debate: why discriminative models can't be universal classifiers — caglar_ee · 2026-09-19