The goal of GeoAI is to enable machines to learn to reason and analyse spatially like a human. However, notwithstanding much progress has already been made, GeoAI is still at an early stage. Traditionally, GeoAI has focused on knowledge-based approaches, which are deeply rooted in geographical information science (GIScience). GIScience is a scientific discipline at the intersection of computer science, social science and natural science that studies geospatial information, addresses how people understand the world and how phenomena of the world can be stored, represented in systems and analysed. As in all fields related to AI, data-driven approaches are becoming increasingly dominant in GeoAI. Within this research unit, we aim to focus on research in both knowledge-based and data-driven GeoAI, and explore the possibilities of combining both into hybrid GeoAI.