AI evidence extraction

Beyond the Prompt: Student Strategies, Ethical Reflections, and Learning with ChatGPT in Computer Science

Kehinde Aruleba, Ismaila Temitayo Sanusi, George Obaido, Blessing Ogbuokiri, Ibomoiye Domor Mienye · Technology Knowledge and Learning · 2025

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

8/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

0.42

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1007/s10758-025-09899-7

Methodology & findings

Study design

Longitudinal qualitative study with mixed data collection: 21 undergraduate computer science students completed five Java-based programming activities, maintained weekly reflective journals over four weeks, and participated in semi-structured interviews..

Sample

N = 21, 1 group

Primary method

Qualitative analysis methods are implied (reflective journals, semi-structured interviews, thematic analysis) but specific statistical or analytical software/procedures are not explicitly stated in the abstract.

Main result

The study found that "students evolved from passive users to active co-creators, developing increasingly refined prompting strategies and critically assessing AI-generated outputs." Most students viewed ChatGPT as valuable for "code structuring, debugging, and explanation," though they identified limitations including "generic responses, overreliance, and concerns around authorship and data privacy."

Reports effect sizes.

Research paradigm

Qualitative/Interpretivist

Author conclusions

The authors conclude that "the study proposes a framework for the pedagogical and institutional integration of GenAI tools that balances personalised support with ethical and critical engagement" and that "implications are offered for computing educators, curriculum designers, and policymakers seeking to embed AI responsibly in computer science education."

Risk of bias

Selection bias: voluntary participation of 21 students may not represent all computer science undergraduates; Temporal bias: four-week study period is relatively short for assessing long-term behavioral changes; Self-selection bias: students willing to maintain reflective journals and participate in interviews may have different attitudes toward AI than non-volunteers; Social desirability bias: students may have reported more ethical reflection in interviews than actual practice; Selection bias: Self-selected sample of 21 undergraduate computer science students; Potential reporting bias in reflective journals and interview responses; Hawthorne effect: Students may modify behavior knowing they are being studied; Small sample size limiting generalizability; Short-term study (four weeks) may not capture long-term effects; Selection bias: Volunteer sample of 21 undergraduate computer science students may not be representative of all CS students; Attrition risk: Four-week longitudinal study with weekly journal maintenance may experience participant dropout; Confirmation bias: Students selected for semi-structured interviews may have been those with stronger experiences or opinions about ChatGPT; Social desirability bias: Students may report more ethical reflections or critical engagement than actual practice due to interview setting; Context specificity: Study limited to Java-based activities at one institution, reducing generalizability

Limitations

The study acknowledges several constraints: it represents "a short-term longitudinal qualitative study perspective" limited to 21 students over four weeks, and the findings are specific to undergraduate computer science students engaged in Java-based activities
The scope does not extend to broader computational disciplines or longer-term impacts.

Open questions raised

The study identifies the need for frameworks balancing personalized AI support with ethical engagement; questions about equitable AI policy in higher education for students with disabilities; and guidance for responsible embedding of AI tools in computer science curricula.
The study identifies the need for research on equitable AI policy in higher education, particularly regarding accessibility benefits for students with disabilities, and calls for frameworks that balance personalized AI support with ethical and critical engagement in computing education.
The authors identify the need for frameworks integrating GenAI tools in higher education, equitable AI policies considering students with disabilities, and further research on responsible AI integration in computer science curricula and institutional policy.

Data: not_statedCode: not_statedExtracted from: pdfAgreement 62%

Explore related topics

Responsible AI Use Human-AI Collaboration Qualitative