Causal Machine Learning & Causal ML Libraries (2026): DoWhy, EconML, CausalML

1. Start Here: Why Causal ML

10 entries

Prediction tells you what will happen; causal ML tells you what to do. Core reading on the shift from correlation to decisions.

01Causal Machine Learning: Moving Beyond Prediction to Real DecisionsScience Spectrum article by Gabriel Nguyen (September 2026) on treatment effects, DAGs, DoWhy, EconML and CausalML. 02Causal inference (Wikipedia)Overview of determining causal effects from data. 03Causality (Wikipedia)General background on cause and effect. 04Ladder of causation: The Book of WhyJudea Pearl's site for the book that introduced association, intervention and counterfactual rungs. 05Judea Pearl: Causal Inference in Statistics, a PrimerCompanion site for Pearl's introductory textbook. 06Counterfactual (Wikipedia)Background on reasoning about what would have happened. 07Rubin causal modelThe potential outcomes framework behind most treatment effect estimators. 08Structural causal modelStructural equations as a way to encode causal assumptions. 09Randomized controlled trialThe gold standard that observational methods try to approximate. 10Correlation does not imply causationThe classic trap, including the ice cream and drowning example.

2. Core Concepts

10 entries

Treatment, outcome, confounders, DAGs and treatment effects. The ideas every causal analysis rests on.

11ConfoundingA variable that influences both treatment and outcome. 12Directed acyclic graphThe graph structure used to draw causal assumptions. 13Causal graphGraphs that show cause and effect between variables. 14Do-calculusRules for turning interventional questions into estimable quantities. 15Average treatment effectATE: the average effect of a treatment across a population. 16Mediation analysisVariables on the path between treatment and outcome. 17Collider (statistics)Why controlling for a common effect introduces bias. 18Simpson's paradoxHow aggregated data can reverse a trend. 19Selection biasSystematic differences between compared groups. 20Causal modelOverview of causal models and when effects can be identified.

3. Estimation Methods

13 entries

From propensity scores to double machine learning. The methods behind the libraries.

21Propensity score matchingMatching or reweighting on the probability of treatment. 22Instrumental variables estimationUsing an instrument when confounders are unmeasured. 23Difference in differencesComparing changes over time between treated and control groups. 24Regression discontinuity designExploiting a cutoff that assigns treatment. 25Synthetic control methodBuilding a weighted comparison unit for a treated unit. 26Uplift modellingPredicting the incremental effect of an action on an individual. 27Double/Debiased Machine Learning (paper)Chernozhukov et al. on cross-fitting and orthogonal scores. 28Generalized Random Forests (paper)Athey, Tibshirani and Wager on forest-based estimation, including causal forests. 29Estimation and Inference of Heterogeneous Treatment Effects using Random ForestsWager and Athey's causal forest paper. 30Metalearners for Estimating Heterogeneous Treatment EffectsKunzel et al. on S-, T- and X-learners. 31Quasi-Oracle Estimation of Heterogeneous Treatment EffectsNie and Wager's R-learner. 32Orthogonal Statistical LearningFoster and Syrgkanis on the theory behind orthogonalized learners. 33Inverse probability weightingReweighting samples to correct for selection into treatment.

4. Causal ML Libraries (Python)

16 entries

Open-source libraries for estimation, uplift and validation. DoWhy, EconML and CausalML are the three most used.

34PyWhyOpen-source ecosystem for causal machine learning. 35DoWhyFour-step workflow: model, identify, estimate and refute. 36DoWhy on GitHubSource, releases and issues for DoWhy. 37EconMLHeterogeneous treatment effects with double ML, causal forests and IV estimators. 38CausalML (Uber)Uplift modelling and treatment effect estimation from Uber. 39CausalML on GitHubSource and releases for Uber's CausalML. 40DoubleMLDouble machine learning for Python and R. 41CausalPyQuasi-experimental methods with Bayesian and OLS models. 42CausalTuneAutomated selection and tuning of causal estimators. 43ChiRhoCausal reasoning on top of Pyro. 44scikit-upliftUplift models and metrics with a scikit-learn style API. 45MeridianGoogle's Bayesian causal marketing mix modelling library. 46PyMCProbabilistic programming used for Bayesian causal models. 47PyroProbabilistic programming on PyTorch. 48pgmpyProbabilistic graphical models and causal inference in Python. 49Responsible AI ToolboxIncludes causal analysis built on EconML.

5. R and Other Libraries

10 entries

Causal inference toolkits outside the Python stack. Useful when your team works in R or needs time-series causal tools.

50grf (Generalized Random Forests)R package for causal forests and related estimators. 51CausalImpactBayesian structural time-series estimates of intervention effects. 52MatchItR package for matching in observational data. 53WeightItR package for propensity score weighting. 54tmleTargeted maximum likelihood estimation in R. 55TigramiteCausal discovery and inference on time series. 56bnlearnBayesian network structure learning in R. 57pcalgCausal structure learning and estimation in R. 58TetradJava causal discovery toolkit from Carnegie Mellon. 59Awesome Causal InferenceCurated list of causal inference resources.

6. Causal Discovery Libraries

8 entries

Learn the graph from data when you do not know it. Always review a discovered graph with domain experts.

60causal-learnPython causal discovery algorithms, including PC and GES. 61dodiscoverPyWhy library for causal discovery. 62pywhy-graphsGraph classes for causal modelling. 63LiNGAMLinear non-Gaussian acyclic models for causal discovery. 64Causal Discovery ToolboxAlgorithms for graph recovery from data. 65NOTEARSContinuous optimisation for DAG structure learning. 66CausicaDeep end-to-end causal inference from Microsoft. 67gCastleHuawei's causal structure learning toolchain.

7. In Production

4 entries

How large companies apply causal inference. Engineering write-ups on uplift, experimentation and root cause analysis.

68Decision Making at NetflixNetflix Tech Blog on experimentation and causal reasoning. 69Microsoft Research: Causal InferenceMicrosoft Research group behind DoWhy and EconML. 70AWS: Root cause analysis with DoWhyUsing DoWhy to find the cause of latency problems. 71causaLensCommercial causal AI platform.

8. Free Books and Courses

9 entries

Learn causal inference with code. Textbooks and courses available online.

72Introduction to Causal Inference (Brady Neal)Free course with lecture videos and notes. 73Causal Inference for the Brave and TruePython handbook for applied causal inference. 74Causal Inference: What If (Hernan and Robins)Free textbook on causal inference. 75Causal Inference: The MixtapeFree online econometrics text with code. 76Applied Causal Inference Powered by ML and AIChernozhukov et al. on causal ML with code in Python and R. 77CausalML BookOnline book version of Applied Causal Inference Powered by ML and AI. 78A Crash Course in Causality (Coursera)Introductory causal inference course. 79PyWhy ExamplesNotebook gallery for DoWhy. 80EconML NotebooksExample notebooks for heterogeneous effects.

Related Pages on This Site

8 entries

More directories on causal AI and machine learning. Companion pages.

81Causal AI ModelsCausal AI models, libraries and discovery methods. 82Causal LLMsCausal reasoning and discovery with language models. 83Graph and Causal Representation LearningGraph and causal representation learning. 84Machine LearningMachine learning overview. 85Machine Learning LibrariesLibraries for machine learning. 86XAI FrameworksExplainable AI frameworks. 87Python Libraries for XAIExplainability libraries in Python. 88Reinforcement LearningReinforcement learning resources.

Frequently Asked Questions

What is causal machine learning?

Causal machine learning estimates the effect of an action on an outcome, not just the likelihood of the outcome. Where standard models learn correlations, causal methods ask whether a treatment such as a discount actually changed behaviour, using assumptions, causal graphs and techniques like propensity scores, instrumental variables and double machine learning.

How is causal ML different from traditional machine learning?

Traditional ML predicts what will happen. Causal ML estimates what would happen if you intervened, and what would have happened otherwise. The first helps you predict; the second helps you decide. Both are useful, and they answer different questions.

What are the main causal ML libraries?

In Python, DoWhy (a model, identify, estimate and refute workflow), EconML (heterogeneous treatment effects) and Uber's CausalML (uplift modelling) are the most used. DoubleML, CausalPy and scikit-uplift are also popular, and grf and MatchIt cover R.

What is the difference between ATE and CATE?

The average treatment effect (ATE) is the average causal effect across a population. The conditional average treatment effect (CATE) is the effect for a subgroup or individual. ATE tells you whether an intervention works; CATE tells you for whom, which matters when you must choose whom to target.

What is a confounder, and why does it matter?

A confounder influences both the treatment and the outcome, creating a false appearance of cause. Ignoring one gives precise-looking but wrong estimates. Drawing a causal graph first, with domain experts, is how analysts decide what to control for and what to leave out.

Can I use machine learning to find the causal graph?

Causal discovery libraries such as causal-learn, LiNGAM and dodiscover can propose graphs from data, but they depend on strong assumptions. Treat their output as a hypothesis and review it with domain knowledge.

Where should a beginner start?

Read the Science Spectrum overview, then work through Brady Neal's free course or the Causal Inference for the Brave and True handbook, and try a small DoWhy example with a refutation test.

How current is this list?

Links were checked in November 2026. Sites such as Netflix, Microsoft and causaLens block automated checks and may need a manual click. This page is a resource list, not statistical or business advice.