Using Instrumentation to Infer Difficulty Levels in Isomorphic Questions

In personalized and adaptive learning, creating exam questions and practice problems that are both unique and challenging for each student helps reduce cheating. A common solution is isomorphic questions, where each question has a similar structure but is different in specific values or details set by the instructor. However, variations in these questions can make some versions unintentionally more difficult than others, which may lead to fairness issues. To address this, the paper proposes methods that offer a reliable framework for managing the difficulty levels of isomorphic assessments, helping instructors to identify and adjust question instances with inconsistent challenge levels. This framework employs instruction count analysis, timebased measurement, and operand size evaluation to infer difficulty levels, providing instructors with the means to create fair and pedagogically sound question sets, benefiting both teaching and learning outcomes.