AI evidence extraction

In ChatGPT they trust: a study of students’ perceptions and misuse of ChatGPT in higher education

Lorraine K. C. Yeung, Daisy Pui Lun Chow, Pak Hang Wong, Sam S. S. Lau · AI and Ethics · 2025

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

7/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

1.31

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1007/s43681-025-00855-w

Methodology & findings

Study design

Qualitative descriptive research approach employing semi-structured interviews with 20 university students in Hong Kong to investigate ChatGPT adoption and misuse patterns.

Main result

The study found that "the majority allow the cognitive artifact to liberate them from burdens of learning" and that "reports from misusers also show more pronounced automation bias and perception of AI agency than non-misusers." Additionally, "among the misusers who are aware of the problem of AI hallucination, some exhibit a paradoxical approach of adoption behavior: they continue to rely on ChatGPT despite noticing that the new burdens of cross-checking its output may be induced."

Research paradigm

Interpretivist/qualitative

Author conclusions

The authors conclude that "this study illuminates how actual cases of misuse and problematic adoption behavior unfold, offering useful directions for education of sensible, responsible and effective use of AI in the higher education context." They further suggest that "this may be associated with the students' perception of ChatGPT as agentic, a factor known to positively shape users' affect-based trust in AI."

Risk of bias

Selection bias - small sample of 20 students from single location (Hong Kong); Potential self-selection bias in interview recruitment; Lack of comparison with non-ChatGPT users; Qualitative approach without quantitative validation; Selection bias: Small sample size (N=20) from a single geographic location (Hong Kong) may not be representative of university students globally; Potential social desirability bias in self-reported adoption behavior during interviews; No indication of inter-rater reliability or coding validation procedures; Geographic and cultural specificity may limit transferability; Selection bias: single institution/geographic location (Hong Kong); Small sample size (n=20) limiting generalizability; Potential self-selection bias in participant recruitment; No mention of inter-coder reliability or reflexivity checks in qualitative analysis

Open questions raised

The study addresses the gap of understanding whether university students actually adopt ChatGPT for enhancement versus misuse for disburdenment, and identifies the need for further education on sensible and responsible AI use in higher education contexts.
The study identifies the need for further understanding of how to educate students for sensible, responsible and effective use of AI in higher education, and calls for investigation into why students continue relying on ChatGPT despite awareness of its limitations such as hallucination and the resulting new burdens of verification.
The study identifies the need for educational interventions to promote sensible, responsible, and effective use of AI in higher education contexts. It suggests further investigation into the relationship between perceived AI agency and adoption behavior.

Data: not_statedCode: not_statedExtracted from: pdfAgreement 68%

Explore related topics

Responsible AI Use Doctoral Training Qualitative