Across a variety of domains, formative feedback is often regarded as beneficial, if not crucial to learning. Yet studies show that this assumption does not always hold true: some types of feedback do not benefit learners. This symposium brings together researchers investigating how feedback can be optimized to maximize potential benefits. The four papers include studies investigating the effectiveness of feedback from various sources including expert, peer and automatically generated feedback in the domains of writing and math. The studies use a variety of methodological approaches including behavioral studies, eye tracking, and data mining. The discussion emanating from the results to be reported during the symposium will focus on how these empirical findings can help to inform feedback delivery in the classroom and how to more effectively design automated feedback.