Student Views on GAI Shape Higher Education, Despite 4.4% Participation Rate

The rapid emergence of generative AI is prompting widespread debate about its role in education, and a new pilot study delves into how university students perceive its use within higher education. William Franz Lamberti, Sunbin Kim, and Samantha Rose Lawrence, all from the College of Science at George Mason University, led the research which explores student opinions on this increasingly prevalent technology. Their work is significant because understanding these attitudes is crucial for educators hoping to navigate the integration of GAI into teaching practices. Despite a modest participation rate, the study offers valuable initial insights and highlights the importance of further investigation with larger groups to fully grasp the student perspective on this transformative tool.

Preliminary results indicate a nuanced understanding of GAI, with students recognising both its potential to enhance learning and concerns regarding academic integrity and the development of critical thinking skills. Researchers designed a seven-question survey, administered before and after students viewed an educational video concerning GAI, then re-administered it to measure changes in responses. Participants were recruited from a total of 68 students across two sections of CDS 130, ensuring representation from core curriculum offerings.

Student participation was entirely voluntary, maintaining ethical research practices. The team used publicly available student registration data, accessed via the GMU Student Registration System, to determine enrollment figures and calculate participation rates. To analyze the low participation rate, the study applied binomial distribution modelling, acknowledging the binary nature of participation. Scientists developed 95% confidence intervals (CI) using the Clopper-Pearson method, implemented through R’s binom.test() function. This statistical approach allowed for a robust comparison with a similar pilot study conducted in Summer 2025, revealing statistically significant differences in participation rates, as evidenced by non-overlapping confidence intervals. The research examined individual survey questions, identifying that approximately 20% showed observable changes in response distributions after video exposure. A total of 3 student responses were collected from an enrolled population of 68, corresponding to an overall response rate of approximately 0.044. Statistical analysis, utilising R’s binom.test(), calculated a 95% confidence interval of (0.009, 0.124) for this rate, revealing significant challenges in engaging students in research concerning generative AI. Comparisons were drawn with a forthcoming Summer 2025 study, which projected a 95% confidence interval of (0.142, 0.429).

The non-overlap of these confidence intervals provides evidence that the participation rates between the two studies differ significantly. Although the low response count limits the strength of conclusions, the research team analysed responses to six key questions, discovering that changes in student opinions were observed in 20% of them. Experiments revealed that, following exposure to the lesson material, no students altered their views on the trustworthiness of generative AI, with the majority maintaining a belief that it provides untrustworthy responses. However, one student shifted from a ‘No’ to a ‘Yes’ response when asked about the acceptability of using generative AI to create art, and another moved from ‘Not Sure’ to agreeing that instructors could use generative AI for any task, demonstrating a shift in perception regarding classroom application.

Further analysis of student responses to questions concerning art created in the style of living artists and the factual reliability of information sources showed no change in opinion after the lesson. Students consistently perceived internet searches as more factual than generative AI results, and books and technology as more reliable than people or visual media. The research successfully gathered preliminary data on student attitudes, establishing a foundation for more comprehensive investigations into this rapidly evolving technological landscape. While acknowledging the challenges of securing broad student participation, the study demonstrates the feasibility of exploring this topic and highlights the potential for instructors to proactively address GAI in their teaching. The findings suggest a need for further research to understand how educational approaches might influence student opinions regarding generative AI. The authors recognise the limitations imposed by a small sample size, preventing statistically significant conclusions at this stage.