This paper proposes a novel method that integrates two approaches to diagnosing student input in step-by-step tasks: model tracing and constraint-based modeling. Model tracing focuses on identifying the sequential problem-solving steps taken by students, while constraint-based modeling supports diagnosing student input even when multiple steps are combined into a single step. The proposed method defines constraints as properties that student input has in common with the steps of a strategy, allowing it to provide diagnostics when students deviate from a strategy even when combining multiple steps. In this study, we explore the design of a multi-step strategy diagnostic system and evaluate these diagnostics. As a proof of concept, we generate diagnostics for an existing dataset (n=2136) containing steps taken by students when solving quadratic equations, and compare them to the system diagnostics performed by two teachers on a sample of students’ strategy applications (n=70) and strategy deviations (n=70). As a result, the system diagnostics are found to be consistent with the teacher’s coding for all 140 student steps.
