<![CDATA[Generative artificial intelligence (AI) has made the assessment of student writing one of the most difficult tasks facing university lecturers. When a text may have been drafted, revised, or polished with AI, lecturers must decide what to assess, how to judge the student's contribution, and how to treat students fairly. This qualitative meta-synthesis examined research on the assessment of AI-assisted student writing to identify the criteria, dilemmas, and fairness concerns that shape lecturers' judgments, with particular attention to the Indonesian higher education context. Drawing on a pool of 99 records and a supplementary set of 24 Indonesian records, the synthesis included 43 studies, 27 of them from Indonesian contexts. Indonesian research has documented educators' perceptions, ethical concerns, and strategies, but no study clearly examined how lecturers grade AI-assisted student writing. Using reciprocal translation, refutational synthesis, and line-of-argument synthesis, the analysis produced three theme clusters. The criteria cluster shows a shift from judging the final product to judging the writing process, the writer's authentic contribution, and the transparency of AI use. The dilemmas cluster captures tensions between detection and trust, between permitted and prohibited use, between legitimate language support for EFL writers and unfair advantage, and between rising expectations and lecturers' limited AI literacy. The fairness cluster highlights the risks of false accusations from unreliable detectors, the stigma that encourages concealment, unequal AI literacy among students, and inconsistency across lecturers and policies. The line of argument is that fair assessment requires a move from detecting AI to assessing the writer's contribution. A process-based assessment workflow and rubric are proposed for Indonesian lecturers, together with a research agenda.]]>
Copyrights © 2025
