Probabilistic Active Goal Recognition

Goal Recognition is the task of inferring the goal of an agent from their action logs. Goal Recognition assumes such action logs are collected by an independent process that is not controlled by the observer. Active Goal Recognition extends Goal Recognition by also assigning the data collection task to the observer. This Ph.D. project will provide a unified probabilistic and decision-theoretic perspective to fundamentally solve the central question: how should an observer act in an environment to actively uncover the goal of the agent?