Procedural Generation of STEM Quizzes

Abstract: Electronic quizzes are used extensively for summative and formative assessment. Current Learning Management Systems (LMS) allow instructors to create quizzes through a Graphical User Interface. Despite having a smooth learning curve, question generation/editing process with such interfaces is often slow and the creation of question variants is mostly limited to random parameters. In this paper we argue that procedural question generation greatly facilitates the task of creating varied, formative, up-to-date, adaptive question banks for STEM quizzes. We present and evaluate a proof-of-concept Python API for script-based question generation, and propose different question design patterns that greatly facilitate question authoring. The API supports questions including mathematical formulas, dynamically generated images and videos, as well as interactive content such as 3D model viewers. Output questions can be imported in major LMS. For basic usage, the required programming skills are minimal. More advanced uses do require some programming knowledge, but at a level that is common in STEM instructors. A side advantage of our system is that the question bank is actually embedded in Python code, making collaboration, version control, and maintenance tasks very easy. We demonstrate the benefits of script-based generation over traditional GUI-based approaches, in terms of question richness, authoring speed and content re-usability.