From production to verification: generative AI, doctoral formation, and the leadership of digital education
TL;DR
It is argued that the leadership of digital education should move beyond broad AI policies toward context-sensitive guidance, verification literacy, transparent disclosure norms, and process-based assessment, including the culminating site of doctoral assessment, the dissertation defense.
Abstract
Generative artificial intelligence is often framed in higher education as a problem of academic integrity, assessment security, or technology adoption. This framing is necessary but insufficient for doctoral education, where writing, reading, coding, synthesizing literature, and interpreting evidence are not merely academic tasks but formative practices through which students become scholars. Based on qualitative interviews with twenty-one doctoral students at a large research university in the United States, this study examines how doctoral students understand and negotiate generative AI in their scholarly work. The study began with students in education and was extended through purposive and snowball recruitment to include students across a range of other disciplines, so that the account would reflect more than one scholarly context; interviews were semi-structured. The findings show that AI functions as an access infrastructure, lowering linguistic barriers for some students and technical barriers for others depending on the demands of their scholarly work. At the same time, students engage in careful boundary work between assistance and authorship, distinguishing grammar support, translation, coding help, and conceptual orientation from intellectual substitution. The analysis further suggests that, among these participants, generative AI is shifting doctoral labor from production toward verification: students' distinctive responsibility increasingly lies in judging the accuracy, legitimacy, ownership, and defensibility of machine-assisted work. Under conditions of policy ambiguity, doctoral students also become primary governors of their own AI use, managing disclosure, caution, verification, and risk. The article argues that the leadership of digital education should move beyond broad AI policies toward context-sensitive guidance, verification literacy, transparent disclosure norms, and process-based assessment, including the culminating site of doctoral assessment, the dissertation defense. These claims are offered as analytic propositions grounded in a single-site interpretive study rather than as generalizable findings. Generative AI has not made doctoral education less necessary; it has made its purposes more urgent.