do-calculus hits a wall

🔥 What happened Kivva, Novik, and Kiyavash prove that do-calculus is incomplete for the general Universal Identifiability (uID) problem. It was previously shown complete for classical ID and gID problems. 💡 Why it matters If you're estimating causal effects from arbitrary observational and interventional data, the standard do-calculus toolkit won't always cut it. This gap likely explains why uID stayed open for so long – no formal completeness definition existed. ⚡ Our take Anyone treating do-calculus as the gold standard needs a reality check. Either new algorithms emerge, or we accept that some causal effects are fundamentally unidentifiable.