Causal Intervention
Overview
It asks what would happen if a variable or mechanism were deliberately changed, rather than merely observed, to estimate effects, test explanations, or choose actions. Common approaches include structural causal models and the do-calculus, randomized or targeted experiments, hard- and soft-intervention causal discovery, and counterfactual or Bayesian optimization methods that select interventions under costs. Applications range from gene perturbation and clinical policies to dynamical systems, robot control, and audits of neural-network features, evaluated with synthetic ground truth and real experimental or time-series data. Key challenges include limited or unknown intervention targets, hidden confounding and feedback, and whether conclusions transfer across environments; observational fit alone often cannot establish what an intervention will do.
Latest papers
96 papers7 papers in the last four weeks, down 22% on the four weeks before. 0.1% of all new papers.
DoTime: A Synthetic Benchmark Generator for Interventional and Counterfactual Time Series
Teacher Knows It Best: Spontaneous Symmetry Breaking and Tipping Points in Networked Langevin Dynamics AI Sycophancy
Amortized Bayesian Causal Discovery of Extended Factor Graphs
Beyond Directed Acyclic Graphs: Causal Zeros and Causal Differential Equations
Interventional Score Geometry for Causal Inference
How Rules Represent Causal Knowledge: Causal Modeling with Probabilistic Logic Programming
Equilibrium Causal Games: Separation, Identification, and the Identifiability of Cyclic Latent States
A cubical formalisation of topos causal models: intervention, forcing, and a contextuality obstruction
--safe flag, with the ordered field discharged at ; type-level sheafification and a directed do-calculus are future work.