Educational technology research and development Образование

Utilizing a GPT-based exam assistant to support practice and assessment in higher education: an exploratory study on learner engagement and perceptions

Frosch, Katharina, Lindauer, Friederike, Winkel, Carmen
Springer Nature · 2026-11-20 · DOI 10.1007/s11423-026-10749-1

Аннотация

Instructors increasingly use chatbots to reduce their workload and to support students with individualized learning opportunities. However, there is still limited knowledge about how students engage with chatbot-based formats for practice and assessment compared to similar human-authored formats. This study examines whether chatbot-supported instruction can achieve similar outcomes with respect to students’ engagement, perceived fairness and perceived usefulness while reducing the workload associated with creating practice and assessment activities. We developed an Exam Assistant (EA) that delivers exam-like open-ended questions, including short case studies, provides immediate feedback, and supports automated scoring with instructor review (human-in-the-loop). The EA was implemented in a real classroom setting with N = 30 undergraduate business students at a German university. Students used the EA for practice and for a graded assignment in the first course phase, while in the second course phase, these activities were delivered via conventional Moodle quizzes. This ensured alignment between practice activities and assessment across both instructional formats . Using learning analytics, chatbot interaction logs from 224 sessions with the EA, and post-assessment surveys, we conducted a descriptive, exploratory comparison of students’ engagement patterns and fairness perceptions. Results show comparably high levels of practice intensity and perceived fairness across both instructional modes. Students demonstrated greater engagement with EA than has typically been reported in prior research on educational chatbots. Nevertheless, students perceived Moodle quizzes as more suitable for future graded assessments, a difference that appears to be driven by question format rather than by fairness concerns. Overall, the findings suggest that thoughtfully chatbot-supported practice and assessment may offer a promising pathway to reduce instructors’ workload without evident disadvantages in student engagement or perceived fairness.

Читать у издателя →