Towards Asynchronous Data Science Invention Activities at Scale

Invention activities are carefully designed problem-solving tasks in which learners are asked to invent solutions to unfamiliar problems prior to being taught the canonical solutions. Invention activities are typically used in the classroom setting. As online education becomes increasingly common, there is a need to adapt Invention activities to the asynchronous nature of many courses. We do so in the context of an introductory undergraduate data science course. Using an online programming environment, students work on the tasks in pairs, without instructor support. We analyze the invention process and outcomes from two Invention activities on the challenging topics of classification and clustering. Detailed analysis of recordings of six student pairs shows how activity design supports insights at three levels: nature of models (e.g., the need to normalize); domain concepts (e.g., types of errors), and procedural solutions (e.g., weighting errors). We describe the activities, their design, and their outcomes.