This talk addresses the challenge of understanding the decision processes behind software used for creating thematic maps. Many existing tools offer suggestions for map types and visual variables, but users often lack insight into why certain suggestions are made. To bridge this gap, the 3D4DT approach is introduced, which uses JSON to represent decision trees and maps these trees into interactive 3D scenes for users to explore. The contributions include a controlled vocabulary to create machine-readable descriptions for decision trees and an interactive 3D implementation. The prototype of this approach is web-based and available on GitHub. It uses Node.js for the server, Vitejs for frontend development, and Three.js, a JavaScript library, for creating interactive 3D scenes. The evaluation involved testing the expressiveness of the controlled vocabulary and the usability of the 3D interactive scene through a user study. Participants interacted with decision trees using both 3D scenes and static websites, and the results indicated that the 3D interactive scene helped users better understand complex decision trees. This work is relevant to developers and users of software for automatic thematic map creation. Developers can encode decision trees as machine-readable data, making their software's decision processes available for reuse. Users benefit from an interactive format to explore how map creation decisions are made. Overall, this approach promotes transparency in algorithmic decision-making for geospatial software and has implications for various open-source geospatial tools that use decision trees.