AI evidence extraction

AI-Assisted Research Writing: Graduate Students' Experiences, Outcomes and Academic Integrity

Leonora Fulgencio De Jesus · International Journal of Learning Teaching and Educational Research · 2026

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

9/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.26803/ijlter.25.5.8

Methodology & findings

Study design

Convergent parallel mixed methods design combining quantitative survey data (Likert scale measures of perceived usefulness, perceived ease of use, and behavioral intention using TAM framework) with qualitative open-ended responses analyzed via inductive coding (process coding and in vivo coding).

Primary method

Convergent parallel mixed methods design; descriptive survey research

Main result

The study found that "ChatGPT was the most popular AI tool, and over three-fourths of them noted that they use it regularly in writing their thesis," with perceived usefulness, perceived ease of use, and behavioral intention to use all rated "to a great extent" (weighted means ranging from 3.75 to 3.90). Additionally, "the results show a strong positive relationship between the researchers' experiences with AI and their writing outcomes, as revealed by the correlation coefficient of .757," indicating that "respondents who report higher levels of experience with AI tools tend to demonstrate better writing outcomes."

Research paradigm

Pragmatist / Mixed methods (convergent parallel design combining quantitative and qualitative strands)

Author conclusions

"Whether AI should be part of graduate research is no longer a question; it has become a reality, and the academy needs to emerge as a force to influence its use fairly. In the coming years, graduate programs should take constructive action by incorporating AI literacy into their research activities, creating clear outlines and contextual policies on the use of AI, and funding professional development in order for faculty members to mentor students on ethical issues of human-AI collaboration in academic writing." The authors also conclude that "AI literacy is already a mandatory component of research competence, the same way information literacy was previously, and graduate programs can no longer afford to disregard it."

Risk of bias

Selection bias: Universal sampling of 96 responsive graduate students from 120 enrolled; 24 non-responders excluded, potentially introducing non-response bias; Attrition/Non-response bias: Study excluded 20% of enrolled population due to non-response during data collection; Self-report bias: Reliance on self-reported survey data regarding AI use and academic integrity practices without objective verification; Temporal confounding: Qualitative sample (n=30) was subset of quantitative sample (n=96), not random selection from full population; Lack of control/comparison group: No control group without AI tool experience for comparison; Correlation vs. causation: Authors acknowledge inability to infer causation from correlational design; prior writing skill, digital literacy, and institutional support not controlled; Selection bias: Universal sampling of 96 responsive students from 120 enrolled (24 non-respondents excluded); non-response bias possible; Self-report bias: Survey-based data on perceived usefulness and behavioral intention; Timing bias: Data collected during thesis writing process, not post-completion; Social desirability bias: Students may overstate ethical adherence when reporting academic integrity practices; Confounding variables: Study does not control for prior writing skills, digital literacy, or institutional support factors; Selection bias: Universal sampling of only 96 responsive students out of 120 enrolled, with 24 non-respondents excluded; Voluntary participation may introduce self-selection bias; Cross-sectional design prevents causal inference; observed correlations may reflect prior writing skill, digital literacy, or institutional support rather than AI experience alone; Single institution context (government higher education institution in Bulacan) limits generalizability; Self-reported survey data subject to social desirability bias regarding academic integrity claims

Open questions raised

Limited empirical data on thesis writing among graduate students in Philippine higher education context
Limited empirical evidence capturing how graduate students describe and practice responsible AI tool use during thesis writing
Fewer studies examining how TAM variables relate to measurable writing performance outcomes specifically among graduate students
Need for further research on disparities in actual AI ethics practices vs. stated ethical intentions in high-stress writing environments
Need for program-specific differences analysis across different graduate disciplines
Local empirical data specifically concerning thesis writing among graduate students in Philippine higher education context remains limited

Extracted from: pdfAgreement 77%

Explore related topics

AI Manuscript Writing Academic Writing Mixed Methods