Item response theory in educational performance assessment: a review of models, challenges, and emerging directions in education

Abstract Performance-based assessment has become central to modern education, particularly in domains that emphasize practical skills, creativity, and problem-solving. Unlike traditional tests, these assessments rely on authentic tasks, complex rubrics, and human judgment, raising concerns about fairness, reliability, and validity. This structured narrative review synthesizes foundational and recent research on Item Response Theory (IRT) in performance- and task-based educational assessment. It critically compares classical dichotomous, polytomous, multidimensional, and many-facet IRT models in terms of their assumptions, scoring requirements, interpretability, and suitability for complex performance data. Particular attention is given to challenges involving local dependence, sample size, calibration, multidimensionality, rater effects, and the fairness of score interpretation. The review also distinguishes established methodological developments from emerging directions: response times, process data, and established adaptive-testing principles already demonstrate ways of extending IRT beyond final task outcomes, whereas artificial intelligence (AI)-supported scoring, adaptive performance assessment, and broader learning-analytics applications require further evidence regarding construct validity, transparency, subgroup fairness, and educational usefulness. Overall, the review shows that greater model complexity does not inherently produce better measurement; the value of IRT depends on aligning model choice with the assessment design, intended interpretation, and educational consequences of score use. This perspective links psychometric rigor with transparent, valid, and equitable educational decision-making.