AI evidence extraction

Constructing Legitimacy in AI-Assisted Academic Writing: Responsibility, Limitation, and Disclosure in Higher Education

Jabreel Asghar · Global Social Sciences Review · 2026

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

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.31703/gssr.2026(xi-i).16

Methodology & findings

Study design

Small-scale exploratory qualitative study using an online questionnaire with both Likert-scale items and open-ended questions.

Sample

N = 25, 3 groups

Primary method

Iterative qualitative coding of open-ended responses. Initial summary of structured Likert-scale responses provided descriptive overview. No inferential statistical methods were employed. Analysis focused on identifying recurring conditions and reasoning patterns through careful reading and thematic synthesis.

Main result

The study found that "across the 25 responses, these judgments were guided by recurring expectations. AI tended to be viewed more favourably when its role stayed limited, when responsibility for interpretation and argument remained clearly with the human writer, and when its use was made" transparent. Legitimacy in AI-assisted academic writing depends on three interconnected conditions: limitation of AI's role, retention of human responsibility, and disclosure of AI involvement.

Reports effect sizes.

Research paradigm

Interpretivist/qualitative

Author conclusions

The authors conclude that "legitimacy does not depend on whether AI is used or not, but on the specific conditions through which responsibility, limitation of AI involvement, and disclosure are upheld in academic practice." They further state that "authorship continues to matter, not because every word must be written by a human hand, but because responsibility for the text must stay visible and clearly attributable to the author." They emphasize that "what is needed is greater clarity regarding the conditions under which AI use can remain acceptable within academic practice."

Risk of bias

Small non-representative sample (n=25); Voluntary participation may introduce self-selection bias; Geographic diversity without stratification details; No mention of inter-rater reliability or coder agreement procedures; No discussion of researcher reflexivity or bias mitigation; Small sample size (n=25) limiting generalizability; Non-representative sampling (convenience/purposive selection); Self-selection bias (voluntary participation); Geographic clustering (participants from South Asia, Middle East, Europe, Central Asia, Africa but distribution not specified); Potential social desirability bias in self-reported judgments about AI use; Selection bias: Participants were self-selected volunteers, not randomly sampled; Sample size: Small sample (n=25) limits generalizability; Geographic/institutional selection bias: Participants came from specific regions (South Asia, Middle East, Europe, Central Asia, Africa) and may not represent all higher education contexts; Lack of representativeness: Authors explicitly state sample was not chosen to be statistically representative

Limitations

The study acknowledges that "the sample was not chosen to be statistically representative," and the findings indicate that "while there was broad agreement on the importance of responsibility and disclosure, the exact boundaries of acceptable use remained less clear." Additionally, "the variation in participants' judgments shows that acceptable use cannot be reduced to a simple set of fixed rules."

Open questions raised

The paper identifies a gap in existing research: "Despite the expanding literature on AI in higher education, much of the existing work remains focused on student behaviour, academic misconduct, or institutional policy. Less attention has been given to how legitimacy is constructed through everyday academic practice." Future research directions include the need for frameworks that clarify how judgments about responsibility and acceptable assistance are made, and greater clarity on acceptable limitations of assistance, the continued importance of human responsibility, and the role of disclosure in academic practice.
Limited attention to how legitimacy is constructed through everyday academic practice (supervision, peer review, manuscript preparation)
Gap between formal policies and actual practitioner judgments in academic settings
Need for greater clarity on conditions under which AI use can remain acceptable within academic practice
Clearer guidance needed on acceptable limitations of assistance, continued importance of human responsibility, and role of disclosure
The paper identifies a gap in existing literature: "Less attention has been given to how legitimacy is constructed through everyday academic practice." The authors note that much discussion focuses on policies and detection tools but neglects how academics actually evaluate writing in daily practice during supervision, peer review, and manuscript preparation.

Extracted from: pdfAgreement 64%

Explore related topics

AI Manuscript Writing Academic Writing Survey