Reliable decision support using counterfactual models

Peter Schulam, Suchi Saria

Research output: Contribution to journalConference article

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.

Original languageEnglish (US)
Pages (from-to)1698-1709
Number of pages12
JournalAdvances in neural information processing systems
Volume2017-December
StatePublished - Jan 1 2017
Event31st Annual Conference on Neural Information Processing Systems, NIPS 2017 - Long Beach, United States
Duration: Dec 4 2017Dec 9 2017

Fingerprint

Supervised learning
Learning algorithms
Decision making
Trajectories
Planning

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Reliable decision support using counterfactual models. / Schulam, Peter; Saria, Suchi.

In: Advances in neural information processing systems, Vol. 2017-December, 01.01.2017, p. 1698-1709.

Research output: Contribution to journalConference article

@article{4f83214da41044de87c2c1fc32fb4d7f,
title = "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.",
author = "Peter Schulam and Suchi Saria",
year = "2017",
month = "1",
day = "1",
language = "English (US)",
volume = "2017-December",
pages = "1698--1709",
journal = "Advances in neural information processing systems",
issn = "1049-5258",

}

TY - JOUR

T1 - Reliable decision support using counterfactual models

AU - Schulam, Peter

AU - Saria, Suchi

PY - 2017/1/1

Y1 - 2017/1/1

N2 - 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.

AB - 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.

UR - http://www.scopus.com/inward/record.url?scp=85046992528&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85046992528&partnerID=8YFLogxK

M3 - Conference article

VL - 2017-December

SP - 1698

EP - 1709

JO - Advances in neural information processing systems

JF - Advances in neural information processing systems

SN - 1049-5258

ER -