Timezone: »

Interpretable Off-Policy Evaluation in Reinforcement Learning by Highlighting Influential Transitions
Omer Gottesman · Joseph Futoma · Yao Liu · Sonali Parbhoo · Leo Celi · Emma Brunskill · Finale Doshi-Velez

Wed Jul 15 12:00 PM -- 12:45 PM & Thu Jul 16 01:00 AM -- 01:45 AM (PDT) @ None #None

Off-policy evaluation in reinforcement learning offers the chance of using observational data to improve future outcomes in domains such as healthcare and education, but safe deployment in high stakes settings requires ways of assessing its validity. Traditional measures such as confidence intervals may be insufficient due to noise, limited data and confounding. In this paper we develop a method that could serve as a hybrid human-AI system, to enable human experts to analyze the validity of policy evaluation estimates. This is accomplished by highlighting observations in the data whose removal will have a large effect on the OPE estimate, and formulating a set of rules for choosing which ones to present to domain experts for validation. We develop methods to compute exactly the influence functions for fitted Q-evaluation with two different function classes: kernel-based and linear least squares, as well as importance sampling methods. Experiments on medical simulations and real-world intensive care unit data demonstrate that our method can be used to identify limitations in the evaluation process and make evaluation more robust.

Author Information

Omer Gottesman (Harvard University)
Joseph Futoma (Harvard University)
Yao Liu (Stanford University)
Sonali Parbhoo (Harvard University)
Leo Celi (MIT)
Emma Brunskill (Stanford University)
Finale Doshi-Velez (Harvard University)

More from the Same Authors