Search papers, labs, and topics across Lattice.
This paper establishes a precise mathematical correspondence between the graphical representation of the do-operator and the functional manipulation of causal mechanisms in deterministic acyclic structural causal models. The authors demonstrate that the dependencies removed through graph surgery align exactly with those eliminated by replacing target mechanisms with constants, formalizing this relationship through a key theorem. This work not only clarifies the interplay between graph theory and causal inference but also provides criteria for when a graph accurately reflects the dependencies of its associated mechanisms.
Replacing causal mechanisms with constants in structural models reveals a surprising equivalence to graph surgery, clarifying how interventions affect dependencies.
The $\operatorname{do}$-operator is described graphically by deleting arrows into its targets and functionally by replacing their mechanisms with constants. To call these operations equivalent is not yet a mathematical statement: one returns a graph and remembers only the targets, whereas the other returns mechanisms and also remembers the imposed values. We make a dependency-level comparison precise for deterministic acyclic structural causal models with finitely many endogenous variables. If $\operatorname{Graph}(F)$ extracts the dependencies of a mechanism family $F$, our main theorem is $\operatorname{Graph}(F^\iota)=\operatorname{Surg}(\operatorname{Graph}(F),T_\iota)$. Thus replacing target mechanisms removes exactly the dependencies removed by graph surgery. For a model $M=(G,F)$ whose graph may contain unused arrows, we characterize when the same equality holds with $G$ in place of $\operatorname{Graph}(F)$; it holds for every intervention exactly when $G$ records the dependencies of $F$ exactly. We then define the intervened model, characterize its run, show how sequential interventions combine, and prove that an outcome depends only on interventions at its actual dependency ancestors.
