A Scoping Review of Generative Artificial Intelligence Boundaries in Educational Assessment Systems

Abstract

Generative artificial intelligence (GenAI) is rapidly entering technology-enabled learning environments, supporting assessment processes such as scoring, feedback, item design, analytic coding, and integrity monitoring. This scoping review mapped GenAI functions in assessment systems, identified boundaries for appropriate use, and examined how human and institutional responsibilities are framed. Peer-reviewed English-language articles published between January 2023 and December 2025 were retrieved from ERIC, Web of Science, and Scopus on January 7, 2026. Of the 1,360 records retrieved, 43 met the inclusion criteria. Analysis examined assessment function, judgment configuration, boundary types, and governance framing. GenAI was most frequently studied as a rubric-applying scorer, with emerging roles in feedback mediation, assessment design, analytic encoding, and integrity monitoring. Across studies, acceptable use was consistently framed as conditional, bounded by validity, reliability, fairness, interpretability, privacy, pedagogical alignment, and governance. Fully autonomous high-stakes use was not supported; hybrid human–AI configurations dominated, positioning reliability and responsibility not merely as technical metrics but as emergent properties of assessment system design.