Reliable Decision Support using Counterfactual Models

Abstract

Decision-makers are faced with the challenge of estimating what is likely to happen when they take an action. For instance, if I choose not to treat this patient, are they likely to die? Practitioners commonly use supervised learning algorithms to fit predictive models that help decision-makers reason about likely future outcomes, but we show that this approach is unreliable, and sometimes even dangerous. The key issue is that supervised learning algorithms are highly sensitive to the policy used to choose actions in the training data, which causes the model to capture relationships that do not generalize. We propose using a different learning objective that predicts counterfactuals instead of predicting outcomes under an existing action policy as in supervised learning. To support decision-making in temporal settings, we introduce the Counterfactual Gaussian Process (CGP) to predict the counterfactual future progression of continuous-time trajectories under sequences of future actions. We demonstrate the benefits of the CGP on two important decision-support tasks: risk prediction and "what if?" reasoning for individualized treatment planning.

Published in the proceedings of Neural Information Processing Systems (NIPS) 2017

Cited by in corpus (32)

• A Review of Challenges and Opportunities in Machine Learning for Health
• Algorithmic recourse under imperfect causal knowledge: a probabilistic approach
• Tutorial: Safe and Reliable Machine Learning
• Can You Trust This Prediction? Auditing Pointwise Reliability After Learning
• Estimating Counterfactual Treatment Outcomes over Time Through Adversarially Balanced Representations
• Calibrating Healthcare AI: Towards Reliable and Interpretable Deep Predictive Models
• Time Series Deconfounder: Estimating Treatment Effects over Time in the Presence of Hidden Confounders
• Preventing Failures Due to Dataset Shift: Learning Predictive Models That Transport
• MultiVerse: Causal Reasoning using Importance Sampling in Probabilistic Programming
• Uncertainty Estimation and Out-of-Distribution Detection for Counterfactual Explanations: Pitfalls and Solutions
• Counterfactual Predictions under Runtime Confounding
• Scaling Gaussian Processes with Derivative Information Using Variational Inference
• Representation Balancing MDPs for Off-Policy Policy Evaluation
• Active Learning for Decision-Making from Imbalanced Observational Data
• Learning When-to-Treat Policies
• I-SPEC: An End-to-End Framework for Learning Transportable, Shift-Stable Models
• Hidden Incentives for Auto-Induced Distributional Shift
• Risk Variance Penalization
• Learning "What-if" Explanations for Sequential Decision-Making
• G-Net: A Deep Learning Approach to G-computation for Counterfactual Outcome Prediction Under Dynamic Treatment Regimes
• Policy Analysis using Synthetic Controls in Continuous-Time
• Neural Pharmacodynamic State Space Modeling
• Estimating Individual Treatment Effects with Time-Varying Confounders
• Sequential Deconfounding for Causal Inference with Unobserved Confounders
• A scoping review of causal methods enabling predictions under hypothetical interventions
• Counterfactual Prediction Under Selective Confounding
• Patient-Specific Effects of Medication Using Latent Force Models with Gaussian Processes
• When the Oracle Misleads: Modeling the Consequences of Using Observable Rather than Potential Outcomes in Risk Assessment Instruments
• The Impact of Time Series Length and Discretization on Longitudinal Causal Estimation Methods
• Cause-Effect Deep Information Bottleneck For Systematically Missing Covariates
• Deep Bayesian Estimation for Dynamic Treatment Regimes with a Long Follow-up Time
• Errors-in-variables Modeling of Personalized Treatment-Response Trajectories