State of the art frameworks for ad hoc teamwork i.e., for enabling an agent to collaborate with others “on the fly”, pursue a data-driven approach, using a large labeled dataset of prior observations to model the behavior of other agents and to determine the ad hoc agent’s behavior. It is often difficult to pursue such an approach in complex domains due to the lack of sufficient training examples and computational resources. In addition, the learned models lack transparency and it is difficult to revise the existing knowledge in response to previously unseen changes. Our prior architecture enabled an ad hoc agent to perform non-monotonic logical reasoning with commonsense domain knowledge and predictive models of other agents’ behavior that are learned from limited examples. In this paper, we enable the ad hoc agent to acquire previously unknown domain knowledge governing actions and change, and to provide relational descriptions as on-demand explanations of its decisions in response to different types of questions. We evaluate the architecture’s knowledge acquisition and explanation generation abilities in two simulated benchmark domains: Fort Attack and Half Field Offense.