Human-Artificial Intelligence (HAI) collaboration in writing offers opportunities to enhance efficiency and booststudent confidence; however, it also carries risks, such as reduced creativity, over-reliance on AI-generatedcontent, and academic integrity (Kim & Lee, 2023

We argue that the decisions writers makeabout whether to use generative AI in their process are, at their core, decisionsnot about platforms but about what writing is for and how writing happens.

Hispresentation, based on students’ written reflections of their experiences usinggenerative AI, highlighted key differences between how students understoodwriting (as a way to gather and transcribe information for outside readers)and how the field talks about writing (as a way to do complex thinking andcreate new ideas).

By bringing in more explicit discussions of writing and learning inour course, we hoped to address the main critique discussed at that panel:students using generative AI did not value the writing process or understandthe relationship between writing and critical thinking.

developing an explicitunderstanding of how to use generative AI ethically, and experimenting withgenerative AI tools in their writing in an approved setting.

We hopethat the assignments, readings, and activities help our students become moreconfident and knowledgeable about their choices to use or not use generativeAI in certain situations.

The course, designed as a semes-ter-long inquiry, invites students to explore how generative AI affects writ-ing and writers.

. We want students to articulate thelimitations of generative AI writing technologies and also develop clear ethicalboundaries for themselves as writers as they make decisions about whether andhow to use generative AI.

“What does it meanto be a writer using AI?” Extending the scope of revision beyond individualassignments gives students the “time and space for further consideration of awriting problem” (Downs 66).

We argue that the decisions writers makeabout whether to use generative AI in their process are, at their core, decisionsnot about platforms but about what writing is for and how writing happens.

This recursivity is intentionaland important, as it helps students practice what “expert readers” do: studentscome back to a text over and over again, revising their understanding of thesource itself as well as their own evolving ideas about how generative AI affectswriters.

We see the course as an exercise in sustained revision: each assignmentcarefully links to the next, and students revise their thinking about writing,generative AI, and their own writing philosophies as they move through theassignment progression.

Hispresentation, based on students’ written reflections of their experiences usinggenerative AI, highlighted key differences between how students understoodwriting (as a way to gather and transcribe information for outside readers)and how the field talks about writing (as a way to do complex thinking andcreate new ideas).

AI collaboration or cheating? Using explainable authorship verification tomeasure AI assistance in academic writing

. GenAI can be misused for UCG, allowing students to delegate entire writing tasks, potentiallybypassing authentic learning. Detecting such misuse is particularly difficult when human and AI inputs areblended, paraphrased, or obfuscated (Weber-Wulff et al., 2023).

These toolscan unfairly penalise non-native English speakers, whose simpler writing styles are often misclassified as AI-generated (Kim & Lee, 2023). Addressing these limitations requires more equitable, transparent methods capableof capturing nuanced human-AI collaboration patterns.

For instance, Gebreegziabher and colleagues (2023) found that higher-performing studentsactively refine AI-generated content, while lower-performing students rely passively on AI, limiting deeperengagement. Similarly, Stanford’s CoAuthor project (Hai, 2023) illustrates productive collaboration practices,demonstrating how AI can enhance, rather than replace, student effort.

he style of a text can be analysedthrough a diverse array of stylistic attributes, including lexical properties (e.g., word usage, sentence length),syntactic traits (e.g., function word use, punctuation), structural layout, content indicators (e.g., n-grams), andidiosyncrasies (e.g., typos or grammar mistakes) (Abbasi & Chen, 2008; Zheng et al., 2006).

). Applications of AI ineducational settings include adaptive learning systems, recommender systems, predictive analytics, chatbots,and, more recently, genAI tools such as large language models (LLMs).

GenAI can be misused for UCG, allowing students to delegate entire writing tasks, potentiallybypassing authentic learning. Detecting such misuse is particularly difficult when human and AI inputs areblended, paraphrased, or obfuscated (Weber-Wulff et al., 2023).

These stylistic markers are instrumental in AV, as they serve to identify an author’s distinctive writing style anddetermine whether a specific text aligns with that style (Koppel & Schler, 2004; Potha & Stamatatos, 2014).

Authorship Verification (AV)—a method traditionally used to determine whether two documents were authoredby the same individual (Koppel & Schler, 2004)—offers a promising yet underexplored solution in educationalcontexts.

Distinguishing AI-generated text from human-authored content is necessary for understanding student learningbehaviours, supporting skill development, and maintaining academic integrity. Analysing student writingpatterns can help educators understand students’ use of AI, track their writing skill progression, and identifysupport needs (Oliveira et al., 2024; Pan et al., 2025)

By integrating lexical and syntacticfeatures into a transparent and robust AV framework, we effectively distinguished between student-authored andAI-generated texts, even under mimicry scenarios.

Distinguishing AI-generated text from human-authored content is necessary for understanding student learningbehaviours, supporting skill development, and maintaining academic integrity. Analysing student writingpatterns can help educators understand students’ use of AI, track their writing skill progression, and identifysupport needs (Oliveira et al., 2024; Pan et al., 2025)

Human-Artificial Intelligence (HAI) collaboration in writing offers opportunities to enhance efficiency and booststudent confidence; however, it also carries risks, such as reduced creativity, over-reliance on AI-generatedcontent, and academic integrity (Kim & Lee, 2023)