Ladder of Causation
Updated 2026-08-15
INTRODUCTION
English translation pending.
CORE DEFINITION
The ladder of causation, due to Judea Pearl, sorts causal questions into three levels: observation, what is seen; intervention, what follows if X is done; and counterfactual, what would have been had X not been done. Higher levels require explicit causal structure to answer, and the level a system can handle is exactly what separates statistical pattern matching from an intelligence that grasps real causes.
SCAFFOLDING EFFECT
Reduce cognitive load
- Grade the capability: place a system or an argument on one of the three levels - Match the evidence: data for observation, experiment for intervention, model for counterfactual - Refuse the mismatch: reject any correlational answer given to a causal question
Anchor fast decisions
Each rung defines what a system can answer at all. Observation supports prediction from joint data, intervention requires a way to manipulate or emulate manipulation, and counterfactuals require a generative model of the world being reasoned about. The higher the level a system handles, the more causal structure it represents, and the more flexible its intelligence becomes.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Causal_modelverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS