Games-Based Learning systems, particularly those that use advances from Intelligent Tutoring Systems (ITS) to provide adaptive feedback and support, have proven potential as learning tools. Taking their lead from commercial games such as Little Big Planet and SuperMarioMaker, these systems are increasingly turning to content creation as a learning activity and to better engage a broader audience. Existing programming puzzle games such as Light-Bot, COPS and Gidget allow users to build their own puzzles within their games. However, open-ended content creation tools like these do not always provide users with appropriate support. Therefore, player-authors may create content that does not embody the core game mechanics or learning objectives of the game. This wastes the time of both the creator and of any future users who engage with their creations. Better content creation tools are needed to enable users to create effective content for educational games. A significant barrier to using user-authored problems in learning games is the lack of expert knowledge about the created content. Many Games-Based Learning systems take cues from ITS and use expert-developed contextual hints and content for individualized support and feedback. Data-driven methods exist to generate hints and estimate which skills particular problems may involve, but these methods require sufficient data collection and knowledge engineering to turn prior student work into hints. No prior work has shown that data-driven methods can be used within a programming game. In addition, data for user-generated levels may be very sparse, so evaluation is needed to determine if these methods can be used in this domain. In this work, I present a set of best practices for designing authoring tools that encourage users to build gameplay affordances into their content. First, I show that requiring users to solve their own levels effectively filters some of the least desirable puzzles, including deliberately impossible or unpleasant levels. Next, I show that the quality of user-authored BOTS puzzles can be improved using level editors designed with these practices. Furthermore, I show that hints and feedback for this content can be created using data-driven methods. I also provide estimates of the amount of data needed to provide adequate hint coverage for a new level using this method. Combined, these findings form an effective framework for integrating user-authored levels into the BOTS game and provide an example of how to build a game from the ground up with user-generated content in mind. These contributions will help future designers create tools that can effectively guide, filter, and leverage user-generated content that will contain gameplay affordances that support the intended game and learning objectives. [The dissertation citations contained here are published with the permission of ProQuest LLC. Further reproduction is prohibited without permission. Copies of dissertations may be obtained by Telephone (800) 1-800-521-0600. Web page: http://www.proquest.com.bibliotheek.ehb.be/en-US/products/dissertations/individuals.shtml.]