AI evidence extraction

Integrating generative artificial intelligence in African higher education: university students’ awareness, attitudes, and use of ChatGPT in Zambia

S Mudenda, Moses Mukosha, Ruth Lindizyani Mfune, B Kathewera, Imukusi Mutanekelwa, B Mwanza et al. · Frontiers in Education · 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

7.33

FWCI

Top 10%

Impact

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.3389/feduc.2026.1814033

Methodology & findings

Study design

Cross-sectional survey study conducted from December 1, 2024, to March 31, 2025, among university students at five institutions in Zambia.

Sample

N = 1829, 10 groups

Primary method

Data entered into Microsoft Excel 2013 and exported to SPSS version 26.0 (IBM Corp., Armonk, NY, USA). Descriptive statistics summarized participant characteristics. Primary outcome measure (ChatGPT perception) assessed using four items coded and summed to generate composite perception score, dichotomized as negative (4-6) or positive (7-8). Sensitivity analyses conducted treating composite attitude score as continuous variable and exploring alternative cut-points. Exploratory Factor Analysis (EFA) with principal component analysis (PCA) performed using Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett's test of sphericity (KMO >0.5, Bartlett's p<0.05 indicating appropriateness). Factors retained based on Kaiser's criterion (eigenvalues >1) and scree plot inspection. Oblique Oblimin rotation applied. Internal consistency assessed using Cronbach's alpha (≥0.70 indicating good reliability). Chi-square (χ²) test examined associations between categorical variables. Variables with p<0.200 in univariate analysis included in multivariable logistic regression. Statistical significance set at p<0.050.

Main result

The study found that "Overall, 96.8% of students had heard of ChatGPT, and 85.6% reported having used it before this study." Additionally, "Respondents aged 40 years and above (adjusted odds ratio [aOR] = 5.91, 95% confidence interval [CI]: 1.23-28.33, p = 0.026) were more likely to have a positive attitude towards ChatGPT than those aged between 18 and 25 years." The research also identified that "Factor 1 was labelled "Academic Risks and Threats" and included eight attitude variables, while Factor 2 was labelled "AI Technology Appeal" and comprised five variables."

Reports effect sizes and confidence intervals.

Research paradigm

Post-positivist

Author conclusions

The authors concluded: "This study provides important insights into university students' awareness, attitudes, and usage patterns of ChatGPT in Zambia. The findings suggest that while many students recognise the potential academic benefits of generative AI technologies, concerns regarding ethical use and academic integrity remain prominent. These results highlight the need for universities to develop clear institutional policies, promote AI literacy, and integrate responsible AI use into higher education curricula."

Risk of bias

Social desirability bias from self-reported data; Selection bias from purposive selection of study sites; Sampling skew toward health-related programs (91.1%); Voluntary participation may have influenced distribution across universities; Cross-sectional design prevents causal inference; Social desirability bias due to self-reported data on academic integrity behaviors; Selection bias from voluntary participation; Sampling bias due to purposive selection of study sites; Sample composition bias: 91.4% of participants in health-related programs; Information loss from dichotomization of continuous measures; Selection bias: Purposive selection of five universities may limit generalizability to all Zambian universities; Social desirability bias: Self-reported data on academic integrity and responsible AI use; Voluntary participation: May have influenced distribution of sample across universities and academic programs; Sample composition bias: 91.4% of sample from health-related programs, limiting transferability to other disciplines; Attrition/response rate: 95% response rate, but those unavailable during collection were excluded; Information loss: Dichotomization of continuous/ordinal measures may reduce statistical power

Limitations

The authors stated: "First, its cross-sectional design only captures a snapshot in time and does not allow for the establishment of causal relationships between the variables." Additionally, "the study relied on self-reported data, which may be subject to social desirability bias, where participants might over-report positive or responsible behaviours related to academic integrity." Furthermore, "the sample was heavily skewed towards students in human health-related programs (91.1%), which may limit the generalizability of the findings to students in other academic fields, including engineering, arts, or business." The authors also noted: "We acknowledge that dichotomising continuous or ordinal measures may lead to loss of information and reduced statistical power."

Open questions raised

Need for longitudinal research designs to understand how perceptions and usage patterns of generative AI technologies evolve over time. Future research should include broader range of academic programs beyond health-related fields. Need for comparative studies assessing quality of student work produced with and without AI-assisted tools. Qualitative approaches to better understand how institutional policies influence student behavior and explore differences in AI adoption across academic disciplines.
The authors identify several research gaps: (1) The need for longitudinal research designs to understand how perceptions and usage patterns of generative AI technologies evolve over time; (2) Research should include a broader range of academic programs beyond health-related fields; (3) Qualitative approaches are needed to better understand patterns of AI use; (4) Examination of how institutional policies influence student behavior; (5) Exploration of differences in AI adoption across academic disciplines; (6) Comparative studies assessing the quality of student work produced with and without AI-assisted tools; (7) The need for a first large-scale multicenter study examining determinants of ChatGPT adoption across Zambian universities.
Limited large-scale multicenter studies examining ChatGPT adoption across Zambian universities
Need for longitudinal research designs to understand how perceptions and usage patterns of generative AI evolve over time
Qualitative approaches needed to better understand adoption patterns
Research examining how institutional policies influence student behavior

Extracted from: pdfAgreement 61%

Explore related topics

AI Literacy Training AI Governance Survey