Causal Identifiability

Follow this topic Ask about this topic

Causal identifiability studies whether observations, interventions, or other information determine a causal graph or quantity uniquely, rather than merely predicting outcomes. Methods include graphical criteria such as do-calculus, structural equation models with assumptions about noise or functional form, and information from proxies, instruments, multiple environments, or targeted interventions. When unique recovery is impossible, partial-identification methods derive bounds, while newer work also examines finite-sample reliability and computational feasibility. A central challenge is that conclusions depend on assumptions and available data: stronger assumptions can make identification possible, but must be justified and tested against alternative models.

Latest papers

128 papers20 papers in the last four weeks, up 54% on the four weeks before. 0.1% of all new papers.

PAC-Causal Learnability: From Identification to Finite-Sample Guarantees
Jul 25, 2025Mian Wei, Somesh Jha, David PageCausal Effect Estimation Causal Identifiability
Duke University · University of Wisconsin-Madison
Clustering and Pruning in Causal Data Fusion
May 21, 2025Otto Tabell, Santtu Tikka, Juha KarvanenClustering Causal Identifiability
Department of Mathematics and Statistics FI-40014 University of Jyvaskyla, Finland
Multi-View Causal Discovery without Non-Gaussianity: Identifiability and Algorithms
Feb 27, 2025Ambroise Heurtebise, Omar Chehab, Pierre Ablin +2Structural Causal Models Causal Discovery
Inria, CEA, University of Paris-Saclay, France · Carnegie Mellon University, Pittsburgh PA, USA · ENSAE, CREST, IP Paris, France +2
Causal Effect Estimation under Networked Interference without Networked Unconfoundedness Assumption
Feb 27, 2025Weilin Chen, Ruichu Cai, Jie Qiao +2Causal Effect Estimation Causal Identifiability
School of Computer Science, Guangdong University of Technology, Guangzhou, 510006, China · School of Computer Science, Guangdong University of Technology, Guangzhou, 510006, China and Peng Cheng Laboratory, Shenzhen, China · Department of Engineering, University of Cambridge, Cambridge CB2 1PZ, United Kingdom
Identifiability Analysis of Linear ODE Systems with Hidden Confounders
Oct 29, 2024Yuanyuan Wang, Biwei Huang, Wei Huang +2Parameter Identifiability Causal Identifiability
The University of Melbourne · University of California, San Diego
Towards Complete Causal Explanation with Expert Knowledge
Jul 10, 2024Aparajithan Venkateswaran, Emilija PerkovićStructural Causal Models Causal Discovery
Microsoft · Redmond, WA 98052, USA · Department of Statistics +2
Causal Discovery in Mixtures of Populations
Nov 13, 2023Bijan Mazaheri, Spencer Gordon, Yuval Rabani +1Latent Variable Models Causal Discovery
Thayer School of Engineering, Dartmouth College, Hanover, New Hampshire, USA · Fragment Data Technologies, San Francisco, California, USA · Department of Computer Science, Hebrew Univeristy of Jerusalem, Jerusalem, Israel +1
Epsilon-Identifiability of Causal Quantities
Jan 27, 2023Ang Li, Scott Mueller, Xin Shu +1Causal Identifiability Causal Inference
Cognitive Systems Laboratory, Department of Computer Science, University of California, Los Angeles, Los Angeles, California, USA.

Jul 25, 2025J/K move · Enter open · S save

PAC-Causal Learnability: From Identification to Finite-Sample Guarantees

Mian Wei, Somesh Jha, David Page

Duke University · University of Wisconsin-Madison

Causal Effect Estimation Causal Identifiability Statistical Learning Theory

Open paper AskPDF arXiv

Causal identification asks whether a causal target can be recovered from the observational distribution in an idealized infinite-data setting. In practice, only finite samples are available, and identifiability alone does not determine how much data are needed for reliable estimation. We introduce Probably Approximately Correct Causal (PACC) learnability, which formalizes uniform accuracy-confidence guarantees over a causal model class. From a PAC-learning perspective, we ask how many samples are sufficient to estimate a causal target to accuracy

ε

with probability at least

1-δ

. We show that PACC sample complexity is shaped by three recurring factors: causal identification strength, target complexity, and nuisance complexity, and derive a general sufficient bound linking them to finite-sample causal learning. We illustrate the framework through self-controlled case series, instrumental variables, and propensity-score-based estimation, showing how distinct causal methods can be analyzed within a common framework.