Focusing on semi-supervised learning on graphs, we discuss a variety of ways to incorporate active learning in this process. These methods involve a human-in-the-loop as part of the learning process and are especially effective for semi-supervised learning with modest amounts of training data. Active learning methods rely on the computer to suggest specific data for the human to provide label information that is used iteratively in the active learning process. We show various ways on how to do this using PDE-based methods on graphs.