Learning Representations for Counterfactual Inference

By Fredrik D. Johansson, Uri Shalit, David A. Sontag. 2016

844 citations recorded by the source corpus.

Observational studies are rising in importance due to the widespread accumulation of data in fields such as healthcare, education, employment and ecology. We consider the task of answering counterfactual questions such as, "Would this patient have lower blood sugar had she received a different medication?". We propose a new algorithmic framework for counterfactual inference which brings together ideas from domain adaptation and representation learning. In addition to a theoretical justification, we perform an empirical comparison with previous approaches to causal inference from observational data. Our deep learning algorithm significantly outperforms the previous state-of-the-art.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Representation & Self-Supervised Learning

Citations in the corpus

Concrete Problems in AI Safety (2016, methodology): “We are excited by recent work applying counterfactual reasoning techniques to machine learning problems [30, 120, 151, 160, 77, 137] though there appears to be much work remaining to be done to scale these to high-dimensional and highly complex settings.”
Causal Effect Inference with Deep Latent-Variable Models (2017, positive): “We follow [25, 48] and use 1000 replications of the simulated outcome, along with the same train/validation/testing splits.”
On the Fairness of Disentangled Representations (2019, background): “Finally, representation learning has recently proved to be useful in counterfactual fairness [55, 45].”
A Meta-Transfer Objective for Learning to Disentangle Causal Mechanisms (2019, background): “Johansson et al. (2016) examine counterfactual inference and formulate it as a domain adaptation problem.”
Woulda, Coulda, Shoulda: Counterfactually-Guided Policy Search (2018, methodology): “Model-based methods for off-policy evaluation have recently been improved to account for the distribution shift between the data-collecting policy and the policy to be evaluated (Johansson et al., 2016; Liu et al., 2018).”
Deep-Treat: Learning Optimal Personalized Treatments From Observational Data Using Neural Networks (2018, methodology): “The approach taken in (Johansson, Shalit, and Sontag 2016) is somewhat different to ours, though, in that they minimize the predictive loss combined with discrepancy between the factual and counterfactual representation populations.”
Deep Neural Networks for Estimation and Inference: Application to Causal Effects and Other Semiparametric Estimands (2018, background): “networks have been argued (experimentally) to outperform the previous state-of-the-art in causal inference (Westreich et al., 2010; Johansson et al., 2016; Shalit et al., 2017; Hartford et al., 2017).”
Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations (2020, methodology): “A large number of methods have been proposed for performing causal inference using observational data in the static setting (Johansson et al., 2016; Shalit et al., 2017; Alaa & van der Schaar, 2017; Li & Fu, 2017; Yoon et al., 2018; Alaa & van der Schaar, 2018; Yao et al., 2018) and only a few…”
Deep Counterfactual Networks with Propensity-Dropout (2017, methodology): “In particular, we compare the MSE (averaged over 100 experiments) achieved by the DCN-PD with those achieved by k nearest neighbor matching ( k -NN), Causal Forests with double-sample trees (Wager & Athey, 2015), Bayesian Additive Regression Trees (BART) (Chipman et al., 2010; Hill, 2012), and Bal-a”
Representation Balancing MDPs for Off-Policy Policy Evaluation (2018, methodology): “102 Our work is inspired by recent research that learns complicated outcome models (reward models in 103 RL) to estimate HTE using new loss functions to account for covariate shift [10, 19, 2, 11].”
The Selective Labels Problem: Evaluating Algorithmic Predictions in the Presence of Unobservables (2017, background): “[16], counterfactual inference techniques can be broadly categorized as: (1) parametric methods which model the relationship between observed features, treatments, and outcomes.”
A Survey of Learning Causality with Data: Problems and Methods (2018, methodology): “Counterfactual Regression [69], the TARnet [124] and the Causal Effect Variational Autoencoder (CEVAE) [84], we are able to learn a representation zu parameterized by neural networks for each instance u based on (xu ,du ,yu ).”
What-If Reasoning with Counterfactual Gaussian Processes (2017, background): “Potential outcomes are often applied to crosssectional data (e.g., Bottou et al. 2013; Johansson et al. 2016), but have also been used to estimate the causal effect of a sequence of actions in discrete time on a final outcome (e.”
From Real‐World Patient Data to Individualized Treatment Effects Using Machine Learning: Current and Future Methods to Address Underlying Challenges (2020, methodology): “Balancing Neural Networks, 20 Treatment-Agnostic Representation Network, 51 Counterfactual Regression, 51 and Similarity Preserved Individual Treatment Effect 57 The ground truth causal effect of the treatment is defined as the difference between receiving the treatment Y ( 1 ) i and not receiving t”