Search papers, labs, and topics across Lattice.
This paper introduces a method for the non-parametric identification of counterfactual queries without the need for a fully specified causal graph by leveraging structural assumptions derived from the queries themselves. By establishing a topological ordering of relevant variables induced by counterfactual inquiries, the authors reduce the identification task to a linear programming problem, enabling the bounding of arbitrary counterfactual queries. The approach generalizes the classical bounding framework of Tian and Pearl (2000) and is validated through case studies that illustrate its practical utility in the absence of complete causal knowledge.
Counterfactual queries can be bounded using a linear programming approach that requires no complete causal graph, revealing insights even with incomplete domain knowledge.
Non-parametric (partial) identification of counterfactual queries typically relies on a fully specified causal graph. Motivated by settings with incomplete domain knowledge, we challenge this requirement by leveraging structural assumptions that are inherently implied by the query itself. We show that any counterfactual inquiry induces a, mostly partial, topological ordering over relevant variables, which, in turn, enables an explicit query parametrisation reducing the identification task to a linear program. This allows bounding arbitrary counterfactual and nested counterfactual queries. Our work can be viewed as a generalisation of the classical bounding framework of Tian and Pearl (2000), originally developed for probabilities of causation. We also prove the \emph{tightness} of our bounds by constructing structural causal models that attain the bounds whilst being compatible with both the observed data and the query-implied order. To assess both the generality and practical utility of the proposed bounding procedure, we revisit several case studies from the literature, demonstrating how the derived bounds can be used to yield informative insights even in the absence of an input causal graph.
