off-policy estimation (OPE) methods allow us to learn and evaluate
decision-making policies from logged data. This makes them an attractive choice
for the offline evaluation of recommender systems, and several recent works
have reported successful adoption of OPE methods to this end. A