Maximilian May

Project

Scalable Oral AI for Higher Education Assessment

Higher education is entering a new assessment era. As generative AI becomes embedded in students’ everyday workflows, written submissions and conventional online exercises are becoming weaker signals of genuine understanding. Maximilian’s project develops an agentic formative assessment system that brings key strengths of oral examinations into a scalable digital format. Through interactive, oral exam-style dialogue, the system probes how students reason, surfaces misconceptions in real time, and provides targeted feedback that helps learners improve throughout a course. The aim is not only to assess answers, but to capture the quality of thought behind them. Scientifically, the project addresses a timely and important question: how should AI-supported assessment systems be designed so that they are pedagogically sound, aligned with course goals, and capable of improving cognitive engagement? The research combines design science with experimental evaluation to iteratively develop an agentic oral examiner and test it in authentic higher-education settings. By examining both assessment quality and downstream learning outcomes, the work contributes new evidence on how conversational AI can support deeper conceptual understanding rather than superficial performance. It also advances broader research on trustworthy agentic systems in education. From a practical perspective, the project responds to a major unmet need in universities and professional education: high-quality individualized assessment that remains feasible in large-scale settings. Oral examinations are widely valued because they reveal reasoning, allow adaptive questioning, and reduce the limits of standardized testing, yet they are rarely deployable at scale due to staffing constraints. A robust AI-based alternative could offer institutions a new layer of continuous assessment, richer learning analytics, and more meaningful performance signals in AI-saturated environments. That creates clear relevance for educational platforms, universities, online programs, and workforce training providers seeking more reliable and engaging ways to evaluate learning.