AI evidence extraction

Algorithmic Mediation and Student Creativity: Large Language Models, Academic Writing, and the Convergence of Ideas in Irish Higher Education

Ziyad Abdulaziz Almeshal, Majed E. Alenazi, Abdulaziz Zaid Albasheer, Saud Rashad Alanazi, Abdulaziz Sami Alsalman · Review Journal of Social Psychology & Social Works · 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/10.71145/rjsp.v4i1.550

Methodology & findings

Study design

Cross-sectional survey design with 150 undergraduate and postgraduate students from three Irish universities (University College Cork, the University of Galway, and Maynooth University).

Sample

N = 150, 11 groups

Primary method

SPSS version 29 was used for quantitative analysis. Descriptive statistics (means, standard deviations, percentage distributions) were calculated for survey items. Inferential analysis included: independent-samples t-tests to examine gender-based differences, one-way ANOVA to compare responses across institutions and disciplinary groups, and Tukey post hoc tests for pairwise contrasts where significant disciplinary differences were identified. Open-ended responses were analyzed thematically following Braun and Clarke stages (familiarisation, initial coding, theme development, review, and refinement).

Main result

The study found that "ChatGPT is widely used and strongly valued for practical reasons" with participants reporting that "it saved time, supported idea development, and helped them engage with difficult concepts." However, the research revealed a more complex pattern: "Many participants reported uncertainty about the originality of AI-assisted ideas, lower confidence in generating ideas without support, and a sense that work produced with ChatGPT may be becoming more similar across users." Notably, "Fewer than half reported a strong sense of ownership over work completed with ChatGPT assistance." Additionally, "Overall, 117 participants (78.0 per cent) selected either agree or strongly agree with the statement that widespread ChatGPT use would make students' ideas look more similar."

Reports effect sizes and confidence intervals.

Research paradigm

Positivist/empiricist with interpretive elements

Author conclusions

"The main contribution of the study lies in showing that this tension is visible not only in theoretical debate or experimental research, but also in students' own accounts of everyday university use. In that sense, the question for higher education is not simply whether generative AI should be adopted or resisted. It is how such tools can be integrated in ways that support learning without reducing opportunities for independent thought and original judgement. A considered response to generative AI therefore requires more than either enthusiasm or prohibition. It requires pedagogical approaches, assessment practices, and institutional guidance that recognise both the value and the limits of algorithmic support."

Risk of bias

Self-report bias and social desirability bias regarding originality, independence, and AI reliance; Recall bias in reporting frequency and patterns of ChatGPT use; Selection bias due to non-representative sampling from three universities only; Temporal confounding: cross-sectional design cannot establish causal effects; Participation incentive may have influenced selection of respondents; Selection bias: stratified quota sampling from three institutions only, not nationally representative; Recall bias: participants reporting on past experiences and perceptions; Social desirability bias: responses about independent thinking and AI reliance may be influenced by perceived social expectations; Measurement bias: self-reported perceptions rather than objective measures of creativity or textual similarity; Temporal bias: cross-sectional design captures single time point, not causal relationships; Selection bias: Stratified quota sampling not statistically representative of all Irish students; Self-report bias: Self-reported perceptions may not correspond to objective measures of creativity or actual writing similarity; Social desirability bias: Participants may underreport or overreport reliance on AI tools when discussing independent thinking; Recall bias: Participants asked to report experiences and frequencies from memory; Volunteer bias: Small participation incentive offered, potentially attracting particular types of participants; Temporal bias: Cross-sectional design captures perceptions at single time point, cannot assess causality

Limitations

The authors state: "First, the cross-sectional design does not permit causal inference
The study captures reported experiences and perceptions at one point in time and cannot establish whether ChatGPT use causes changes in creativity or originality
Second, the study relies on self-reported data
Participants' perceptions of creativity, dependency, or ideational similarity may not correspond directly to observable performance
Third, although the sample includes students from three universities and a range of disciplinary fields, it was not designed to be statistically representative of the wider student population in Ireland
Fourth, responses may have been influenced by recall bias or social desirability, especially in relation to independent thinking, originality, and reliance on AI tools."

Open questions raised

Longitudinal work to determine whether perceptions of dependency, ownership, or similarity change over time
Experimental studies and text-based analyses to examine whether perceived convergence corresponds to measurable similarity in written outputs
Investigation of whether disciplinary culture, prior digital familiarity, or prompting skill influence how students use AI
Research on AI literacy and teacher preparation, especially where institutional support is uneven
Comparative work across different AI systems to clarify whether patterns are specific to ChatGPT or reflect broader large language model features
Much existing literature focuses on usability, ethics, policy, and literacy rather than creativity and ideational convergence as specific research problems

Extracted from: pdfAgreement 69%

Explore related topics

AI Manuscript Writing Academic Writing Survey