Generative-AI use and study habits of college students in fully remote, hybrid and fully onsite courses: a simulated comparative study
Abstract
Simulated cross-sectional survey of 612 undergraduates comparing weekly generative-AI use and study habits across fully remote, hybrid and fully onsite course delivery, and their relation to final grade.
Full text
Introduction
Universities moved large parts of teaching online during the pandemic and have kept many hybrid formats since [1]. At the same time, generative AI assistants became freely available to students [2]. Remote learners work with less external structure, so they may lean more on AI tools for explanation and drafting [3]. Whether that reliance helps or harms learning is an open question [4,5]. We ask three questions: how does AI use differ across remote, hybrid and onsite delivery; how do study habits differ; and how are both related to course outcomes?
Methods
Design and participants
We simulated a cross-sectional survey in one semester. Participants were 612 undergraduates taking the same three introductory courses delivered in three formats. Group sizes are given in Table 1. Students self-reported demographics, weekly hours of AI tool use, planning habit (a mean of five items on a 1–5 scale), study sessions per week and procrastination (1–5). Final course grade was taken from the registry.
Statistical analysis
Group differences were tested with one-way analysis of variance and described with eta-squared. Pairwise differences use Welch intervals. Grade was modelled by ordinary least squares with delivery mode, AI use, AI use squared and planning habit; onsite delivery was the reference. All data were generated with a fixed random seed, and the analysis script is available from the corresponding author.
Results
Participants
Table 1 summarises the three groups.
| Delivery mode | n | AI use, h/week | Planning (1–5) | Sessions/week | Final grade, % |
|---|---|---|---|---|---|
| Fully remote | 204 | 4.2 (2.2) | 3.37 (0.66) | 5.2 (1.4) | 71.1 (7.6) |
| Hybrid | 211 | 2.9 (1.7) | 3.56 (0.65) | 6.3 (1.4) | 75.2 (7.2) |
| Fully onsite | 197 | 2.1 (1.3) | 3.57 (0.64) | 6.0 (1.5) | 74.0 (7.6) |
Table 1. Participants and outcomes by delivery mode. Values are mean (standard deviation).
AI use by delivery mode
Weekly AI use differed by delivery mode (F = 67.0, eta-squared = 0.18; Figure 1). Remote students used AI 2.0 h/week more than onsite students (95% CI 1.7 to 2.4).
Study habits
Planning habit differed by delivery mode (F = 6.1, eta-squared = 0.02), with remote students lowest and hybrid and onsite students similar. Hybrid students reported the most study sessions per week, and remote students reported the most procrastination (Figure 2).
AI use and course grade
Grade rose slightly with AI use in the raw data, and the fitted curves for the three modes were similar in shape but offset in height (Figure 3). After adjustment the association was not distinguishable from zero.
Table 2 and Figure 4 give the adjusted model. Relative to onsite delivery, remote students scored about three points lower, and the hybrid difference was small. The coefficients for AI use and AI use squared both had confidence intervals that included zero.
| Term | Coefficient | 95% CI |
|---|---|---|
| Intercept (onsite, reference) | 62.71 | 59.21 to 66.21 |
| Fully remote | -3.02 | -4.61 to -1.44 |
| Hybrid | 0.90 | -0.53 to 2.33 |
| AI use (h/week) | 0.42 | -0.41 to 1.25 |
| AI use squared | -0.01 | -0.10 to 0.07 |
| Planning habit (1–5) | 2.94 | 2.06 to 3.83 |
Table 2. Linear model of final grade (R-squared = 0.12).
Discussion
In this simulated sample, remote delivery went with heavier AI use, weaker planning and more procrastination, while hybrid students studied in the most sessions. The lack of a clear adjusted association between AI use and grade is consistent with the view that the benefit of AI depends on how it is used alongside regular study rather than on how much it is used [4,5]. Because the study is cross-sectional and the data are simulated, no causal claim is made. A real study would need randomisation or longitudinal follow-up, validated habit scales and pre-registered analyses [6].
Limitations
The data are simulated, so the effects reflect the assumptions used to generate them. Self-reported AI use is subject to recall error, and students were not randomised to delivery mode.
Data availability
All data are simulated. The generating script and a CSV of the simulated dataset are available from the corresponding author on request.
Ethics statement
No human participants were involved. All records are synthetic.
References
- Bozkurt A, Sharma RC. Emergency remote teaching in a time of global crisis due to coronavirus pandemic. Asian Journal of Distance Education. 2020;15(1):i-vi.
- Kasneci E, Sessler K, Kuchemann S, et al. ChatGPT for good? On opportunities and challenges of large language models for education. Learning and Individual Differences. 2023;103:102274. doi:10.1016/j.lindif.2023.102274
- Broadbent J, Poon WL. Self-regulated learning strategies and academic achievement in online higher education learning environments: a systematic review. The Internet and Higher Education. 2015;27:1-13. doi:10.1016/j.iheduc.2015.04.007
- Zawacki-Richter O, Marin VI, Bond M, Gouverneur F. Systematic review of research on artificial intelligence applications in higher education. International Journal of Educational Technology in Higher Education. 2019;16:39. doi:10.1186/s41239-019-0171-0
- Means B, Toyama Y, Murphy R, Bakia M, Jones K. Evaluation of evidence-based practices in online learning: a meta-analysis and review of online learning studies. US Department of Education; 2010.
- Nosek BA, Ebersole CR, DeHaven AC, Mellor DT. The preregistration revolution. Proceedings of the National Academy of Sciences. 2018;115(11):2600-2606. doi:10.1073/pnas.1708274114
Data & code
All data are simulated; the generating script is available on request.
AI-use disclosure
None
Competing interests and funding
None Funding: None
Open peer review
Accepted with thanks.
Careful, reproducible study. Careful, reproducible study. Careful, reproducible study. Careful, reproducible study.
Careful, reproducible study. Careful, reproducible study. Careful, reproducible study. Careful, reproducible study.
Metrics & integrity
- Views
- 12
- Downloads
- 5
- Integrity audit
- Passed integrity audit (pass), signed off by a human integrity editor on 2026-11-16
- Reviews published
- 2