AI evidence extraction

From integrity to identity: course-level generative AI governance and scholarly subject formation in graduate education

Evelyn Wu · Frontiers in Education · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

2/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3389/feduc.2026.1827251

Methodology & findings

Study design

Qualitative bounded interpretive case study design integrating two complementary data sources: (1) systematic analysis of twenty-two graduate course syllabi collected from courses taught between 2023-2025, coded for governance rationalities; (2) eleven semi-structured interviews with doctoral students (purposively sampled for variation across program area, doctoral stage, and linguistic background).

Sample

N = 33, 5 groups

Primary method

The study employed qualitative analysis methods rather than quantitative statistical testing. Analysis procedures included: (1) Two-cycle inductive and theory-informed coding of syllabi with focused coding of governance rationalities and interpretive analysis of discourse; (2) Reflexive thematic analysis of interview transcripts with line-by-line initial coding, second-cycle clustering into broader thematic categories, and theory-informed interpretation; (3) Iterative analytic memo-writing to document emerging codes, interpretive decisions, and theoretical connections; (4) Cross-source integration through comparative matrices linking syllabus patterns with interview themes. No statistical software or hypothesis testing was employed.

Main result

The study found that "course-level GenAI governance in graduate education operates as a process of scholarly subject formation" where "GenAI governance in this case was not experienced as a single institutional policy. Instead, it appeared through fragmented course-level expectations, including prohibition, conditional authorization, assistive-generative boundaries, pedagogical integration, and silence." Across the syllabi analyzed, four patterned governance regimes emerged, and graduate students responded to this fragmented environment through self-regulation, boundary work, and reputational caution.

Reports effect sizes.

Research paradigm

Interpretivist/Critical qualitative

Author conclusions

"The study makes three contributions. First, it shows that course-level syllabi function as governance artifacts that translate institutional uncertainty into local rules about AI, writing, authorship, and academic integrity. Second, it demonstrates that fragmented governance creates interpretive burdens for students, who must navigate shifting expectations across instructors, assignments, and courses. Third, it advances a theoretical account of GenAI governance as a process of scholarly subject formation, linking policy enactment, governmentality, symbolic power, linguistic capital, and boundary work." The authors conclude that "the governance of GenAI is also the governance of emerging scholarly identity" and that "the challenge, therefore, is not simply to decide whether AI should be banned or allowed. It is to develop governance that protects intellectual agency, supports equitable participation, and helps emerging scholars use new tools critically, ethically, and responsibly."

Risk of bias

Selection bias: Purposive sampling of syllabi and interviews may not represent full range of AI governance approaches; Participant bias: Students who agreed to participate may have stronger opinions about AI governance than non-participants; Social desirability bias: Self-reported AI practices may be shaped by participants' willingness to discuss sensitive topic and desire to present themselves favorably; Memory bias: Interview data rely on retrospective accounts of AI use and policy interpretation; Single-case bias: One college within one U.S. research university limits generalizability across institutional contexts; Researcher positionality: Researcher's familiarity with debates about GenAI and academic integrity may introduce confirmation bias; Selection bias: Purposive sampling of students may have enrolled more reflective or opinionated participants; Potential social desirability bias: Participants may underreport or misrepresent AI use due to sensitivity surrounding academic integrity; Self-report bias: Reliance on recalled accounts of AI practices and perceptions; Researcher positionality: Author's familiarity with higher education debates may influence interpretation; Sample composition bias: Nine of eleven participants are multilingual, potentially overrepresenting this perspective; Attrition/non-response bias: Unknown characteristics of students who declined to participate; Selection bias: purposively selected sample of students who agreed to participate may have stronger opinions about AI governance; Social desirability bias: participants may underreport or misrepresent sensitive AI use practices; Recall bias: self-reported accounts of AI use shaped by memory; Single-author interpretive study: potential for confirmation bias despite reflexivity measures

Open questions raised

Comparative studies across colleges, disciplines, or institutional types to examine whether similar governance regimes appear in STEM, humanities, professional schools, or community colleges
Instructor interviews to clarify how faculty design AI policies and understand authorship, assessment, and student responsibility
Longitudinal research examining how graduate students' AI practices and scholarly identities change over time
Larger studies investigating whether multilingual students experience AI governance differently from native English-speaking students across institutional contexts
Classroom observations and instructor interviews to clarify how AI governance evolves over time and how students' scholarly identities develop in relation to changing technological norms
Research across different national contexts and institutional types beyond single U.S. research university

Extracted from: pdfAgreement 65%

Explore related topics

Institutional Policy AI Governance Case Study