Causal discovery

  • Aims to infer the underlying causal structure (often a Directed Acyclic Graph) from data using conditional independencies, scores, or stability assumptions.
  • A set of algorithms that attempt to uncover causal graphs from observational data using conditional independence, score-based optimization, or asymmetries.

Overlap: Informs Adjusted Methods & Quasi-experiments (by discovering adjustment sets), supports Causal ML (by identifying variable roles), and relates to Alternative Paradigms when standard graph assumptions are relaxed

Key feature: Infers causal structure

Usage: For learning causal DAGs, discovering/detecting exposure pathways


Table of contents