A Lecture-Specific AI-Based Tutor for Higher Education: Pedagogical Design and Empirical Evaluation

Generative AI tools are increasingly used in higher education, yet most available systems lack pedagogical grounding, course alignment, and insight into student learning. This paper presents the development, implementation, and evaluation of the bioTutor, an open-source, course-specific AI chatbot designed to support constructivist learning in large university classrooms. The system integrates a curated knowledge base, a didactically structured interaction design, and a learning analytics dashboard for instructors that summarizes anonymized student-chatbot conversations. To assess students’ perceptions of usefulness, ease of use, learning relevance, output quality, and result demonstrability, we developed an education-adapted extension of the Technology Acceptance Model (edTAM) and applied it in an introductory biology course with 407 enrolled students. During a 23-week deployment, students generated more than 10,000 interactions across over 1000 conversations. Questionnaire data indicate high usability, strong perceived usefulness, and broad interest in adopting similar tools. Usage patterns show that the bioTutor was employed both for learning and exam preparation. These findings suggest that students perceived pedagogically structured, course-grounded AI chatbots as useful for learning and exam preparation, while the lecturer dashboard provided aggregated insights into students’ questions and recurring difficulties. The open-source framework enables adaptation to other disciplines and provides a scalable foundation for further research on didactically informed AI systems in higher education.