AI evidence extraction

Graduate students and AI: Insights into academic writing practices

Berkay Aktepe, Seher Çetinkaya · Education Mind · 2025

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/10.58583/em.4.2.2

Methodology & findings

Study design

Basic qualitative research design using content analysis.

Sample

N = 10, 5 groups

Primary method

Content analysis as the primary analytical method. The authors state: "The qualitative data obtained in this study were analyzed using content analysis. Content analysis is not merely a process of classifying texts; it is an in-depth analytic approach that uncovers latent layers of meaning in participants' statements." Two independent coders performed coding with consensus-building for disputed items to enhance reliability.

Main result

The findings reveal that "graduate students use AI tools in a multidimensional and systematic manner." Specifically, "all participants reported using ChatGPT, indicating that this tool plays an active role in academic writing processes," with students using AI tools "for literature review, idea generation, structuring the conceptual framework, and language refinement." Additionally, "participants emphasized that AI supports idea development, enhances creativity, facilitates academic productivity, and saves time," though "concerns remain regarding originality, ethical use, excessive dependence, and weakening of researcher identity."

Reports effect sizes.

Research paradigm

Qualitative interpretivist

Author conclusions

The authors conclude that "AI tools function as supportive assistants but should not replace human intellectual contribution." They state: "This study demonstrates that graduate students actively and purposefully use AI tools throughout the academic writing process. AI tools enhance productivity, support learning, facilitate literature review, and reduce the cognitive effort required in academic writing. However, concerns remain regarding originality, ethical use, excessive dependence, and weakening of researcher identity." The authors recommend explicit institutional guidelines centering on "the principle that AI is only a tool" while allowing supportive uses but strictly prohibiting actions such as "generating original data (scientific fraud)."

Risk of bias

Selection bias: Convenience sampling from a single institution limits representativeness; Attrition: One participant's data excluded due to insufficient content; Response mode bias: Mix of face-to-face interviews and written forms may influence response depth; Participant self-selection: Only those willing to participate included; Selection bias: Convenience sampling of easily accessible participants; Single institution: Data from one university only; Sample size: Only 11 participants (10 analyzed), reducing representativeness; Data collection method bias: Mixed modes (interviews vs. written forms) may have influenced response depth; Data exclusion: One participant excluded for insufficient content; Researcher positionality not explicitly stated; Selection bias: Convenience sampling of 11 graduate students from a single institution (Ordu University) limits representativeness to broader graduate student populations; Attrition: One participant's data excluded due to insufficient content, reducing sample from 11 to 10; Method bias: Mixed data collection methods (face-to-face interviews vs. written forms) may produce differential response patterns; Small sample size (n=10) limits generalizability and statistical power; Self-report bias: Participants self-report their AI tool usage; no independent verification provided

Limitations

The study employed convenience sampling of 11 graduate students from two programs (primary education and early childhood education) at a single university, which limits generalizability
The authors note that "methodological literature indicates that different data collection modes may influence the depth and form of responses," and one participant's data were excluded "due to insufficient content," reducing the effective sample to 10 participants
The study's context-specific nature (Turkish institution with TÜBİTAK guidelines) and small sample size constrain broader applicability.

Open questions raised

The authors identify that "little is known about how these tools are used, how students benefit from them, and what thoughts or reflections emerge throughout this process." They note the need for future research on regulating AI use in academic writing, stating that this "should not rely solely on technical detection mechanisms but should instead be grounded in ethical awareness, transparent disclosure practices, and clear institutional guidelines."
The authors identified the need to understand which stages of academic writing AI tools are used for and for what purposes, stating: "Thus, it is crucial to investigate the types of AI tools preferred by graduate students and the purposes for which they use them." They also note that "little is known about how these tools are used, how students benefit from them, and what thoughts or reflections emerge throughout this process." Future research directions include developing comprehensive institutional guidelines on AI use and investigating long-term effects on researcher identity and critical thinking.
The authors identified the need for 'explicit guidelines on how AI tools should be used in academic writing processes' and recommend that guidelines 'should center on the principle that AI is only a tool, as defined by the TÜBİTAK Guide (2025); while allowing supportive uses such as grammar correction, they should strictly prohibit actions such as generating original data (scientific fraud) and uploading confidential data to publicly available AI tools.'

Extracted from: pdfAgreement 63%

Explore related topics

AI Manuscript Writing Academic Writing Qualitative