Research Proposal: long-term LLM use on academic writing

What?

This assignment involved designing and presenting a full research proposal examining the long term effects of large language model use on students’ independent academic writing. The proposal required identifying a genuine gap in the literature, synthesising existing research, and translating that gap into a feasible, ethical, and methodologically sound study.

Rather than focusing on whether AI improves writing in the moment, the proposal was deliberately framed around learning transfer. The central question asked whether sustained use of LLMs supports or weakens students’ ability to write independently once AI support is removed. This required designing a study that separated short term assistance from longer term skill development, which proved more challenging than I initially anticipated.

The proposal brought together multiple components developed across the module, including critical literature review, research design, ethical governance, and assessment strategy. It also required clear justification for methodological choices, particularly around avoiding AI detection tools and prioritising developmental approaches to academic integrity.

So what?

The most significant learning from this assignment was realising how easily research questions can default to measuring convenience rather than learning. Many existing studies focus on performance while AI is present because it is easier to observe and quantify. Designing a proposal that tested what students could do without AI forced me to think much more carefully about what meaningful evidence of learning actually looks like.

The literature review highlighted a clear pattern. While short term gains in fluency and organisation are well established, there is far weaker evidence around independence, critical thinking, and originality over time. Engaging critically with this gap shifted my thinking from seeing it as a limitation of existing tools to recognising it as a methodological problem that required a different research design.

Ethics also became central rather than procedural. Decisions to minimise data collection, avoid surveillance based AI detection, and separate research participation from assessment were not simply compliance measures. They directly shaped the type of knowledge the study could produce. This reinforced the idea that ethical design is inseparable from methodological quality, particularly in research involving learning, power, and evaluation.

Developing separate evaluation frameworks for academic reasoning and language proficiency was another key insight. It challenged my earlier assumption that improved writing necessarily indicates improved thinking. This distinction now feels fundamental to how I interpret both AI assisted work and student performance more broadly.

Now what?

This proposal has directly influenced how I approach research moving forward. As I enter the dissertation and computing project module, I am more deliberate about aligning research questions, methods, and evaluation criteria from the outset, rather than allowing the method to dictate the question.

The project has also shaped how I think about AI governance in practice. Moving away from detection and enforcement toward transparency, task design, and reflective use feels both ethically defensible and educationally sound. This perspective aligns closely with my professional context, where scalable, fair approaches to assessment and feedback are critical.

More broadly, the proposal has strengthened my confidence in engaging with contested, evolving topics. Rather than avoiding uncertainty, I now see it as a productive space for research, provided it is approached with methodological care and ethical clarity. This mindset will carry forward into both my academic work and my professional decision making around data, assessment, and emerging technologies.

References

• Almoraie, A. and Alhejaili, M. (2025). ‘A systematic literature review to implement large language models in education’, Education and Information Technologies. https://doi.org/10.1007/s44217-025-00424-7
• Bittle, K. and El-Gayar, O. (2025). ‘Generative AI and academic integrity in higher education: A systematic review and research agenda’, Information, 16(4), p. 296. https://doi.org/10.3390/info16040296
• Bretag, T. and Mahmud, S. (2016). ‘A conceptual framework for implementing exemplary academic integrity policy in Australian higher education’, in Bretag, T. (ed.) Handbook of Academic Integrity. Singapore: Springer, pp. 463–480. https://doi.org/10.1007/978-981-287-098-8_24
• Chan, C.K.Y. (2025). ‘Students’ perceptions of “AI-giarism”: Investigating changes in understandings of academic misconduct’, Education and Information Technologies, 30, pp. 8087–8108. https://doi.org/10.1007/s10639-024-13151-7
• Dai, H., Zhu, X., Shen, S. and Li, Y. (2023). ‘AI-generated feedback on writing: Insights into efficacy and ENL student writing’, International Journal of Educational Technology in Higher Education, 20(1), p. 58. https://doi.org/10.1186/s41239-023-00425-2
• Fan, X., Li, R., Zhang, Y. and Wang, Y. (2024). ‘Comparing the quality of human and ChatGPT feedback on students’ writing’, Journal of Writing Research. https://doi.org/10.1016/j.jowr.2024.100021
• Gibbs, G. (1988). Learning by Doing: A Guide to Teaching and Learning Methods. Oxford: Oxford Polytechnic.
• Howard, R.M. (1995). ‘Plagiarisms, authorships, and the academic death penalty’, College English, 57(7), pp. 788–806. https://doi.org/10.2307/378403
• Kasneci, E. et al. (2023). ‘ChatGPT for good? On opportunities and challenges of large language models for education’, Learning and Individual Differences, 104, 102274. https://doi.org/10.1016/j.lindif.2023.102274
• Lund, B.D. et al. (2026). ‘Student perceptions of AI-assisted writing and academic integrity’, AI in Education, 1(1), p. 2. https://doi.org/10.3390/aieduc1010002
• Pagaling, H., Cochrane, A. and Philp, J. (2024). ‘Academic integrity or academic misconduct? Conceptual and practical tensions in higher education’, Higher Education Research and Development, 43(8), pp. 1564–1584. https://doi.org/10.1080/07294360.2024.2339833
• Ramírez-Montoya, M.S. and Lugo-Ocando, D. (2024). ‘The impact of large language models on higher education’, Frontiers in Education, 9, 1392091. https://doi.org/10.3389/feduc.2024.1392091
• Shen, S. (2023). ‘Application of large language models in the field of education’, International Journal of Emerging Technologies in Learning, 18(4), pp. 24–35. https://doi.org/10.3991/ijet.v18i04.38017
• Susnjak, T. and McIntosh, T.R. (2024). ‘ChatGPT: The end of online exam integrity?’, Education Sciences, 14(6), 656. https://doi.org/10.3390/educsci14060656
• Tai, M. et al. (2024). ‘Impact of ChatGPT on ESL students’ academic writing skills’, Smart Learning Environments, 11, 2. https://doi.org/10.1186/s40561-024-00295-9
• Xu, J. and Zhou, Y. (2023). ‘The impact of AI writing tools on the content and organisation of EFL student writing’, Cogent Education, 10(1), 2236469. https://doi.org/10.1080/2331186X.2023.2236469
• Zawacki-Richter, O. and Marín, V.I. (2024). ‘ChatGPT in higher education: A synthesis of the literature and a bibliometric analysis’, Educational Technology Research and Development. https://doi.org/10.1007/s11423-024-10334-1