Transparency challenges in policy evaluation with causal machine learning: improving usability and accountability
Published in Data & Policy, 2024
This discusses issues around transparency when using causal forest models. It covers both the traditional concerns in machine learning around accountability around letting black-box models make decisions that affect people's lives, and also the need for transparency so decision-makers can actually understand the insights in the causal forest model. It argues that given models are largely not making decisions directly, the greater concern is that models are transparent enough the be usable. It discusses some techniques that could be used to address transparency issues.