Interpretable Personalized Experimentation
This addresses the need for interpretable and maintainable personalized experimentation systems in large-scale production environments like Meta, though it appears incremental as it builds on existing HTE models.
The paper tackles the problem of black-box heterogeneous treatment effect (HTE) models being difficult to understand and maintain in production by presenting a scalable, interpretable personalized experimentation system deployed at Meta, which learns explanations for these models and generates interpretable personalized policies.
Black-box heterogeneous treatment effect (HTE) models are increasingly being used to create personalized policies that assign individuals to their optimal treatments. However, they are difficult to understand, and can be burdensome to maintain in a production environment. In this paper, we present a scalable, interpretable personalized experimentation system, implemented and deployed in production at Meta. The system works in a multiple treatment, multiple outcome setting typical at Meta to: (1) learn explanations for black-box HTE models; (2) generate interpretable personalized policies. We evaluate the methods used in the system on publicly available data and Meta use cases, and discuss lessons learnt during the development of the system.