While there’s much excitement about machine learning, the technology driving our daily interactions with these models are often blackboxed and inaccessible to the broader audience. A major problem is the lack of concrete visual representations that any user can manipulate to understand their results. We aim to provide such a representation by coupling the data and model with a visual-rule based interface. We present Learner’s Permit, an interactive programming game accessible to a nontechnical audience. The object of the game is to teach driving skills to a simulated self-driving car, which allows users to ground their machine learning knowledge in a relevant scenario. We aim to support the entire workflow of machine learning, including data collection and cleaning, training, model choice and model criticism, testing and debugging, and user interface. By combining the principles of visual rule programming and interpretable AI, we hope to build a game that makes machine learning accessible for all audiences through visual rules.

I’m excited to participate in SuperUROP because I want to pursue a piece of research and explore academia. I am interested in both machine learning and game design; with this project, I hope to combine my experience in both subjects to produce an ML game that’s understandable by all. I’m most excited to design, build, and evaluate a new piece of technology.