Authentic assessment in an AI-integrated qualitative research methods course

Abstract

Abstract Authentic assessment has been a recognized aspiration in higher education for nearly four decades, yet its implementation at the undergraduate level has remained constrained by the procedural demands of professional-level work. Generative artificial intelligence reshapes the implementation calculus by lowering the entry barriers that previously confined authentic assessment to capstone courses, professional programs, and elite institutional contexts. This paper presents a single-semester case study of an AI-integrated qualitative research methods course taught to undergraduates in a child and adolescent development program at a regional public university. Twenty-one students used a custom AI class companion without restriction to produce full-length qualitative research manuscripts. The analytic backbone is a five-dimensional model of authentic assessment, with its physical-context dimension extended to require AI inclusion in fields where professional practice now incorporates the technology. Three rubric-proximal proxy indicators were estimated across the corpus: methods specificity, claim-evidence coupling, and theory-interpretation linkage density. The methods specificity proxy compresses toward the upper bound of its scale, indicating that AI scaffolding supports structural elements of qualitative writing effectively. The two judgment-dependent proxies retain wide dispersion. Process tracing of four contrasting cases identifies three pathways that distinguish stronger from weaker outcomes: criteria-based orchestration, irreducible input supply, and substantive iteration. The boundary-condition pattern indicates that authentic assessment in AI-integrated courses remains diagnostic where it needs to be and is supported where it needed support.