Scalable Contrastive Causal Discovery under Unknown Soft Interventions
This work addresses the problem of causal discovery for researchers and practitioners dealing with observational and interventional data, particularly in scenarios with unknown soft interventions, which is an incremental improvement in the field.
The authors tackled the problem of causal discovery under unknown soft interventions and achieved improved causal structure recovery, with their model asymptotically recovering the corresponding identifiable PDAG. Experiments demonstrated scalability to larger graphs and generalization to unseen graphs.
Observational causal discovery is only identifiable up to the Markov equivalence class. While interventions can reduce this ambiguity, in practice interventions are often soft with multiple unknown targets. In many realistic scenarios, only a single intervention regime is observed. We propose a scalable causal discovery model for paired observational and interventional settings with shared underlying causal structure and unknown soft interventions. The model aggregates subset-level PDAGs and applies contrastive cross-regime orientation rules to construct a globally consistent maximal PDAG under Meek closure, enabling generalization to both in-distribution and out-of-distribution settings. Theoretically, we prove that our model is sound with respect to a restricted $Ψ$ equivalence class induced solely by the information available in the subset-restricted setting. We further show that the model asymptotically recovers the corresponding identifiable PDAG and can orient additional edges compared to non-contrastive subset-restricted methods. Experiments on synthetic data demonstrate improved causal structure recovery, generalization to unseen graphs with held-out causal mechanisms, and scalability to larger graphs, with ablations supporting the theoretical results.