The educational data mining community has long acknowledged the "challenge of interpretability" that has grown alongside the adoption of complex machine learning algorithms for educational purposes. Researchers have focused on a variety of approaches for addressing this concern, often turning to methods borrowed from the broader eXplainable AI (XAI) community. However, serious limitations with existing methods have led to calls for a re-imagining of what explainability should look like. The HEXED workshop aims to bring together a community of researchers who can work together to (1) develop a shared vision and common vocabulary for XAI in education, (2) share and disseminate work, (3) create robust methods for increasing interpretability, and (4) develop evaluation metrics for assessing explanations and model interpretability. We propose to achieve this through collaborative sense-making, research poster presentations, and lively discussions surrounding the current and future needs of the community.