From Scaffolding to Substitution? Generative AI, Self-Regulation, and Learner Agency in EFL Academic Writing in Benin
This mixed-methods study of EFL students in Benin reveals that the educational value of generative AI in academic writing depends not on the number of tools used, but on the intensity of use and the learner's ability to self-regulate the technology as a scaffold rather than a substitute for their own effort.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
In university classrooms around the world, a quiet revolution has reshaped how students write. For decades, teachers have watched learners struggle with grammar, vocabulary, and the daunting task of organizing their thoughts into clear arguments. Now, a new kind of digital assistant has arrived. These tools, known as generative artificial intelligence, can instantly correct a sentence, suggest a better word, or even draft an entire paragraph. For students learning English as a foreign language, these tools offer a lifeline, removing barriers that once kept them from participating in academic life. But this convenience brings a difficult question: when a machine helps so much, who is actually doing the writing? Is the tool acting as a helpful tutor, offering temporary support to help a student learn, or has it become a substitute, doing the work the student should be doing for themselves? This tension lies at the heart of a new study from Benin, which moves beyond simply asking if students use these tools to investigate how they use them, and what that usage means for their own ability to think and write.
Researchers Coffi Martinien Zounhin Toboula, Innocent Goudjinou Sakpoliba, and Emmanuel Ohôô Ezin set out to understand this dynamic among undergraduate students at the University of Abomey-Calavi. They gathered data from 186 students and 24 English lecturers, asking not just which tools were popular, but how deeply students relied on them at different stages of writing. The team looked at the entire process, from the initial brainstorming of ideas to the final polishing of the text. They wanted to see if using these tools helped students feel more confident and write better, or if it led to overreliance, where students stopped thinking for themselves. Crucially, they also asked students to report on their own self-control: were they planning their writing, checking the AI's suggestions, and taking responsibility for the final result, or were they simply accepting whatever the machine produced?
The study revealed that the most popular tool by far was ChatGPT, used by 86 percent of the students, followed by other platforms like Grammarly and Gemini. However, the researchers found something surprising about how students interacted with these tools. It did not matter how many different tools a student used; a student who used six different apps did not report any greater benefit than a student who used just one. What mattered was how often they used the tool and how deeply they engaged with it. Students who used AI frequently reported feeling much more supported in their writing, more confident in their abilities, and more skilled at managing their own learning process. Yet, this increased engagement came with a twist. The students who felt the most benefit were also the ones who expressed the strongest concerns about overreliance, the difficulty of verifying facts, and the risk of losing their own voice in the writing.
When the researchers looked at where in the writing process the AI was most helpful, a clear pattern emerged. Students relied on these tools most heavily during the revising and proofreading stages. They used AI to fix grammar, check vocabulary, and polish the final text. While the tools were also used for brainstorming and drafting, the support was significantly stronger when the work was already done and needed cleaning up. The lecturers interviewed in the study confirmed this observation. Nearly all of them saw AI as a powerful aid for fixing language errors and helping students revise their work. But they also voiced serious worries. A large majority of the lecturers expressed concern that students might become too dependent on the technology, potentially handing over the intellectual work of writing to the machine. They noted that while the final text might look perfect, it was becoming harder to tell if the student truly understood the ideas or could write independently.
The most striking finding of the study is that benefit and risk are not opposites. In many discussions about new technology, people assume that if something is helpful, it must be safe, or if it is risky, it must be unhelpful. This study suggests a more complex reality. The students who got the most out of AI were also the most aware of its dangers. They were the ones who used the tools frequently, felt more confident, and practiced better self-regulation, yet they were simultaneously the most worried about the ethical and practical pitfalls of relying on the technology. This suggests that as students become more experienced with AI, they do not simply become blind to the risks; instead, they become more sophisticated in their understanding of both the power and the limits of the tool. The lecturers echoed this sentiment, seeing AI as a valuable scaffold that helps students reach higher, but also as a potential trap that could replace the student's own effort if not managed carefully.
The researchers emphasize that their work does not prove that AI makes students better writers in an objective sense, as they did not grade the actual essays. Instead, the study maps out the landscape of how these tools are being used and perceived. It shows that the number of tools a student uses is less important than the intensity of their engagement. It highlights that the greatest support happens after the initial writing is done, during the revision phase. And it reveals that the boundary between helpful assistance and harmful substitution is not a fixed line, but a delicate balance that depends on whether the student remains in control. For teachers and students alike, the lesson is that the value of generative AI depends entirely on how it is regulated. If students continue to plan, monitor, and evaluate the work, the tool acts as a powerful scaffold. If they stop thinking and simply accept the output, the tool becomes a substitute. The future of writing in the age of AI, the study suggests, belongs to those who can use the machine without letting it take the wheel.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.