Efficient Real-world Testing of Causal Decision Making via Bayesian Experimental Design for Contextual Optimisation

TL;DR

This paper presents a Bayesian experimental design framework for efficient, data-driven evaluation and improvement of personalized treatment decisions in real-world settings, avoiding sub-optimal assignments.

Contribution

It introduces a model-agnostic, information-based Bayesian design method for evaluating and optimizing contextual treatment policies, applicable to both discrete and continuous treatments.

Findings

Outperforms baseline methods in simulation studies.

Reduces the need for sub-optimal treatment assignments.

Effectively evaluates regret of past decisions.

Abstract

The real-world testing of decisions made using causal machine learning models is an essential prerequisite for their successful application. We focus on evaluating and improving contextual treatment assignment decisions: these are personalised treatments applied to e.g. customers, each with their own contextual information, with the aim of maximising a reward. In this paper we introduce a model-agnostic framework for gathering data to evaluate and improve contextual decision making through Bayesian Experimental Design. Specifically, our method is used for the data-efficient evaluation of the regret of past treatment assignments. Unlike approaches such as A/B testing, our method avoids assigning treatments that are known to be highly sub-optimal, whilst engaging in some exploration to gather pertinent information. We achieve this by introducing an information-based design objective, which we optimise end-to-end. Our method applies to discrete and continuous treatments. Comparing our information-theoretic approach to baselines in several simulation studies demonstrates the superior performance of our proposed approach.