Modern practitioners use machine learning to guide their decisions in applications ranging from marketing and targeting to policy making and clinical trials. Such decisions often affect which data

Modern practitioners use machine learning to guide their decisions in applications ranging from marketing and targeting to policy making and clinical trials. Such decisions often affect which data one gets to collect, necessitating efficient planning of experiments/trials or task assignments to human annotators. To embrace automated decision-making, it is crucial to make decisions safely and efficiently, as poor decisions waste human and physical resources, resulting in social and economic costs. To this end, I propose to build a comprehensive research program on closed-loop decision problems with an emphasis on humans in the loop. Specifically, I will investigate novel ways of interacting with humans in realistic settings, develop mathematical algorithms and theory for complex models, and apply them to salient problems in biology, psychology, and economics.