Reality-based tasks for competency-based education: The need for an integrated analysis of subject-specific, linguistic, and contextual task features

In evaluating competency-based education, effective test instruments must address real-life complexities. The impact of subject-specific, linguistic, and contextual task features, alongside central personal characteristics, on the empirical challenge of such tasks is unclear. We developed mathematics tasks from 30 real-world contexts, each with three questions of varying complexity, administered through a systematically rotated experimental design to 535 German grades 9 and 10 students. Various student variables were collected. Generalized linear mixed models revealed that contextual and mathematical task features significantly contributed to task difficulty variance. Language features had no intermediate-level influence, while students' mathematical self-efficacy moderated low task context familiarity's impact. These findings guide the construction of reality-based mathematics tasks to tailor empirical difficulty.