The DOI matches its Crossref record; 46 of the 47 references it cites were found in a registry or an index; 8 of the 23 sources its claims cite were checked for support.
The Effects of GitHub Copilot on Computing Students' Programming Effectiveness, Efficiency, and Processes in Brownfield Coding Tasks
Open accessPublished version of a preprint
Abstract
When graduates of computing degree programs enter the software industry, they will most likely join teams working on legacy code bases developed by people other than themselves. In these so-called brownfield software development settings, generative artificial intelligence (GenAI) coding assistants like GitHub Copilot are rapidly transforming software development practices, yet the impact of GenAI on student programmers performing brownfield development tasks remains underexplored. This paper investigates how GitHub Copilot influences undergraduate students' programming performance, behaviors, and understanding when completing brownfield programming tasks in which they add new code to an unfamiliar code base. We conducted a controlled experiment in which 10 undergraduate computer science students completed highly similar brownfield development tasks with and without Copilot in a legacy web application. Using a mixed-methods approach combining performance analysis, behavioral analysis, and exit interviews, we found that students completed tasks 35% faster (p < 0.05) and made 50% more solution progress p (< 0.05) when using Copilot. Moreover, our analysis revealed that, when using Copilot, students spent 11% less time manually writing code (p < 0.05), and 12% less time conducting web searches (p < 0.05), providing evidence of a fundamental shift in how they engaged in programming. In exit interviews, students reported concerns about not understanding how or why Copilot suggestions work. This research suggests the need for computing educators to develop new pedagogical approaches that leverage GenAI assistants' benefits while fostering reflection on how and why GenAI suggestions address brownfield programming tasks. Complete study results and analysis are presented at https://ghcopilot-icer.github.io/.
Overview
CoChat Summary
TL;DR
In a within-subjects experiment with 10 undergraduate computing students, GitHub Copilot improved brownfield programming task efficiency and solution progress while shifting students from manual coding and web search toward prompting and evaluating Copilot responses. Students reported productivity gains but also concerns about not understanding how or why Copilot suggestions worked.
Key findings
- Students completed Task 1 in 34.9% less time with Copilot (M = 837.0 s, SD = 496.1) than without it (M = 1285.9 s, SD = 503.5), a significant difference (W = 7.0, p = 0.04, r = 0.66).
- Students passed more automated tests with Copilot (M = 6.30, SD = 1.57) than without it (M = 4.20, SD = 1.87), a significant difference (W = 1.50, p = 0.01, r = 0.80).
- Programming activity distributions differed significantly (χ² = 42.01, df = 9, p = 0.004, Cramer’s V = 0.18): time writing code fell from 30.45% to 19.82%, viewing web content fell from 13.51% to 1.91%, while writing prompts rose from 0.00% to 14.27% and viewing Copilot responses rose from 0.00% to 12.45%.
- In code-writing behavior, manual entry dropped from 95.6% without Copilot to 52.5% with Copilot; with Copilot, students pasted Copilot responses 21.3%, inserted using Copilot 9.3%, accepted suggestions 6.6%, and modified suggestions 3.7% of coding time. Higher performers were more selective; low performers spent significantly more code-writing time directly pasting code from Copilot chat (p < 0.001), and top performers showed 83.9% less pasting of Copilot code.
Methods
- Within-subjects, mixed-methods experiment with two conditions: No Copilot (control) followed by Copilot (experimental); treatment order was not counterbalanced, but task order was counterbalanced.
- Participants were 10 undergraduate computer science students at Oregon State University (9 male, 1 female, mean age 26.7), screened for prior web development experience and little to no prior GenAI assistant experience.
- Students completed two highly similar brownfield feature implementations (“Add Distance” and “Add Picture”) in a 3,818-line legacy web application, with automated test suites and up to 30 minutes per feature.
- Data included task completion time, tests passed, second-by-second video coding of programming behaviors, and exit interviews; analyses used non-parametric tests, permutation-based chi-square tests, Markov transition analysis, and Krippendorff’s alpha of 0.91 for primary categories and 0.85 for secondary categories.
Limitations
- The paper states that participants did not all use the same version of the VS Code GitHub Copilot extension because auto-updating was not disabled; “Copilot Edits” appeared during data collection, and one participant used it during 40% of their code-writing time.
- The paper states that the within-subjects design did not counterbalance treatment order, leaving a possible learning effect from the No Copilot treatment unmitigated.
- The paper states that behavioral coding relied on cursor movement as a proxy for attention shifts, and eye tracking would have been more reliable.
- The paper states that the sample was small (10 participants from a single university), focused on web programming with one GenAI tool and limited tasks, and restricted to students with high academic performance in web development coursework, limiting generalizability.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
5 Valid · 0 Not valid · 11 Not checked · 79 Own results and interpretations
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“This growing body of evidence underscores the importance of metacognitive capabilities-specifically, an ability to maintain agency and critical judgment when presented with external solutions-as essential to the effective use of GenAI for programming.”
What the sources say
- Partly supportedcites [23] [23] Dastyni Loksa, Andrew J Ko, Will Jernigan, Alannah Oleson, Christopher J Mendez, and Margaret M Burnett. 2016. Programming, problem solving, and selfawareness: effects of explicit guidance. In Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems . ACM, 1449-1461. - sourceExperimental — accuracy not yet measured
The Introduction says metacognitive awareness matters for learning to code, and the Discussion and Future Work section links recognizing effective and ineffective strategies to more independent problem solving. The paper studies introductory programming with instructional prompts and hints; it does not establish that maintaining agency and critical judgment over GenAI-generated solutions is essential to using GenAI effectively. (LLM, conf=0.86)
- Not checked yetcites [8] [8] Allan Collins, John Seely Brown, and Susan E Newman. 1989. Cognitive apprenticeship: Teaching the crafts of reading, writing, and mathematics. Knowing, learning, and instruction: Essays in honor of Robert Glaser 18 (1989), 32-42. -
- Not checked — no full text of the sourcecites [32] [32] James Prather, Brent N Reeves, Juho Leinonen, Stephen MacNeil, Arisoa S Randrianasolo, Brett A Becker, Bailey Kimmel, Jared Wright, and Ben Briggs. 2024. The widening gap: The benefits and harms of generative ai for novice programmers. In Proceedings of the 2024 ACM Conference on International Computing Education Research-Volume 1 . 469-486. - source
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“This transition reflects what Liang et al. [19] call the”
What the sources say
- Partly supportedcites [19] [19] Jenny T. Liang, Chenyang Yang, and Brad A. Myers. 2023. Understanding the Usability of AI Programming Assistants. ArXiv abs/2303.17125 (2023). https: //api.semanticscholar.org/CorpusID:263870561 - sourceExperimental — accuracy not yet measured
Split into 2 atoms — 0 confirmed, 2 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“(b) students' programming processes differed markedly with and without Copilot;”
boostedstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“- With Copilot, participants spent significantly less time writing code.”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Participants reported mixed experiences regarding how Copilot affected their brownfield coding activities.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
50 more claims
Create a free account to see all 95 claims and to verify the 3 references still open.
- Probably not a substantial claim. 25
- It may state two separate claims as one. 18
- The quote only partly states this claim. 2
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Observational data further reveal reductions in independent debugging and problem decomposition–the very practices thought to cultivate durable expertise (Shihab et al., 2025).”
From Tool, tutor, or crutch?: A grounded theory of cognitive scaffolding and offloading in AI-assisted programming education (2026) · cites this paper as Shihab et al., 2025
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The Boilerplate Blindspot observed here also aligns with industry reports that routine scaffolding from AI can speed delivery on “brownfield” tasks while concealing subtle faults in glue code (Shihab et al., 2025).”
From Tool, tutor, or crutch?: A grounded theory of cognitive scaffolding and offloading in AI-assisted programming education (2026) · cites this paper as Shihab et al., 2025
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Artificial intelligence (AI)-assisted coding is increasingly common in modern research and industry workflows because it significantly improves efficiency, accessibility, productivity, and happiness (1, 2).”
From Learning R with generative AI in a metagenomic data science course (2026) · cites this paper as [2]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“this nuance helps interpret why reported productivity gains on maintenance tasks (Shihab et al., 2025) can co-exist with latent defects in glue code.”
From Tool, tutor, or crutch?: A grounded theory of cognitive scaffolding and offloading in AI-assisted programming education (2026) · cites this paper as Shihab et al., 2025
Citations
Cites
The control of the false discovery rate in multiple testing under dependency
Benjamini, Y., & Yekutieli, D. (2001) · The Annals of Statistics · doi:10.1214/aos/1013699998 · Cited by 10,490
Sweller, J. (2011) · Psychology of Learning and Motivation · doi:10.1016/b978-0-12-387691-1.00002-8 · Cited by 4,856
Gemini: A Family of Highly Capable Multimodal Models
Gemini Team, Anil, R., Borgeaud, S., Alayrac, J.-B., Yu, J., Soricut, R., Schalkwyk, J., Dai, A. M., Hauth, A., Millican, K., et al. (2023) · arXiv · doi:10.48550/arxiv.2312.11805 · Cited by 990
Permutation Tests for Complex Data: Theory, Applications and Software
Pesarin, F., & Salmaso, L. (2010) · Wiley Series in Probability and Statistics · doi:10.1002/9780470689516 · Cited by 515
The Metacognitive Demands and Opportunities of Generative AI
Tankelevitch, L., Kewenig, V., Simkute, A., Scott, A. E., Sarkar, A., Sellen, A., & Rintel, S. (2024) · Proceedings of the CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642902 · Cited by 493
The Robots Are Coming: Exploring the Implications of OpenAI Codex on Introductory Programming
Finnie-Ansley, J., Denny, P., Becker, B. A., Luxton-Reilly, A., & Prather, J. (2022) · Proceedings of the 24th Australasian Computing Education Conference · doi:10.1145/3511861.3511863 · Cited by 463
GitHub Copilot AI pair programmer: Asset or Liability?
Moradi Dakhel, A., Majdinasab, V., Nikanjam, A., Khomh, F., Desmarais, M. C., & Jiang, Z. M. (. (2023) · Journal of Systems and Software · doi:10.1016/j.jss.2023.111734 · Cited by 417
GitHub Copilot AI pair programmer: Asset or Liability?
Dakhel, A. M., Majdinasab, V., Nikanjam, A., Khomh, F., Desmarais, M. C., Ming, Z., & Jiang (2022) · arXiv · doi:10.48550/arxiv.2206.15331 · Cited by 417
Yu, H. (2023) · Frontiers in Psychology · doi:10.3389/fpsyg.2023.1181712 · Cited by 409
The Impact of AI on Developer Productivity: Evidence from GitHub Copilot
Peng, S., Kalliamvakou, E., Cihon, P., & Demirer, M. (2023) · arXiv · doi:10.48550/arxiv.2302.06590 · Cited by 287
Research works printed but not linked: 2
- Microsoft. 2015. Visual Studio Code. https://code.visualstudio.com/
- Tianyi Xu Pranav Vaithilingam Xinyu Li Yuntao Tian Laura Dabbish and Brad A Myers. 2022. A theory-driven approach to AI-assisted programming. CHI Conference on Human Factors in Computing Systems (2022) 1–15.
Cited by
Qiao, Y., Shihab, M. I. H., & Hundhausen, C. D. (2025) · ACM Transactions on Computing Education · doi:10.1145/3785366 · Cited by 31
“The emergence of Large Language Models (LLMs) in artificial intelligence has led to a proliferation in so-called Generative AI (GenAI) assistants (e.g. GitHub Copilot [28]) in both the software industry and computer science education [22, 24, 26, 27, 35, 57, 70, 78, 80, 82].”
Self-citationExploring Student-AI Interactions in Vibe Coding
Geng, F., Shah, A., Li, H., Mulla, N., Swanson, S., Raj, G. S., Zingaro, D., & Porter, L. (2026) · Australasian Computing Education Conference · doi:10.1145/3786228.3786236 · Cited by 51
Liu, D., Fan, G., & Pan, L. (2026) · International Journal of STEM Education · doi:10.1186/s40594-025-00592-w · Cited by 16
Evolution of Programmers' Trust in Generative AI Programming Assistants
Shah, A., Rexin, T., Tomson, E., Griswold, W. G., Porter, L., & Raj, A. G. S. (2025) · Koli Calling International Conference on Computing Education Research · doi:10.1145/3769994.3770029 · Cited by 13
“Ourstudyaimstounderstand how and why studenttrustchanges as they learn about and use GitHub Copilot.”
Jia, Y., Li, C., Chan, A., & Li, Q. (2025) · Communications in computer and information science · doi:10.1007/978-981-95-4499-8_32 · Cited by 3
Metadata
- Authors
- Shihab, M. I. H., Hundhausen, C., Tariq, A., Haque, S., Qiao, Y., & Mulanda, B. W.
- Affiliations
- Oregon State University (US)
- ORCID
- Venue
- Proceedings of the 2025 ACM Conference on International Computing Education Research V.1
- Publisher
- ACM
- Published
- 2025-08-02
- Pages
- 407-420
- OpenAlex
- W4412781211
- Cited by
- 28
- References
- 28
- Citations by year
- 2026: 22 · 2025: 6
- FWCI
- 40.82
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- ACM Conference on International Computing Education Research (ICER) (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University) (cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Science Applications › Teaching and Learning Programming
- Physical Sciences › Computer Science › Information Systems › Software Engineering Research
- Social Sciences › Education › Artificial Intelligence in Education
- Keywords
- GitHub Copilot, undergraduate programming education, programming performance, programming efficiency, programmer behavior, AI-assisted programming, student perceptions of AI
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: The Effects of GitHub Copilot on Computing Students' Programming Effectiveness, Efficiency, and Processes in Brownfield Coding Tasks. https://app.cochat.ai/papers/10.1145/3702652.3744219
- The paper
- Shihab, M. I. H., Hundhausen, C., Tariq, A., Haque, S., Qiao, Y., & Mulanda, B. W. (2025). The Effects of GitHub Copilot on Computing Students' Programming Effectiveness, Efficiency, and Processes in Brownfield Coding Tasks. Proceedings of the 2025 ACM Conference on International Computing Education Research V.1. https://doi.org/10.1145/3702652.3744219
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.