What does the availability of GenAI do to trust between students and teachers? Currently reading Luo (2024)

Cheating is not a new phenomenon by any means, but with the raise of GenAI, it seems to be easier than ever, and teachers seem to be more suspicious than ever. But in contrast to direct plagiarism where a source can easily be identified, the use of GenAI to produce text in assignments can mostly only be suspected*, not proven (unless there is some direct observation of text being copied from a LLM into the document, or if teachers do weird tricks like hiding prompts in white font in pdfs, hoping that students will copy the whole text into a LLM without reading what they are putting in or taking out), because AI detection services are not only not reliable, but also biased (for example more likely to suspect language learners of cheating than native writers). At the same time, students cannot prove innocence when suspected of using LLMs, either. Being wrongly accused, or even just being worried that being wrongly accused might happen, is really not a good feeling, and it is easy to see how trust in all directions would suffer. Luo (2024) explores “How does GenAI affect trust in teacher-student relationships? Insights from students’ assessment experiences“, which I am summarising below.

I have written about the importance of trust between teachers and students before. In this study, Luo (2024) conducts brief online interviews to collect background information, then invites students to a concept mapping exercise, which they then discuss in a follow-up interview. They find four main themes of how students describe trust-building with their teachers in relation to GenAI:

• Navigating the use of AI: In order to minimise the risk of accidentally using GenAI tools in ways that might be judged as cheating, some students refuse to use any AI tool in their assignments, or they use AI detection services before handing in their assignments to check nothing is flagged as potentially AI generated. If there is no risk, no trust is needed.
• Navigating the acknowledgement of AI: Disclosure happens very strategically since it is often not clear how teachers will include the information in their marking. Will it put people who disclose their use at a disadvantage relative to those who did not use AI (or did not disclose it and weren’t suspected of using it)?
• Navigating interpersonal relationship building: When students feel like the teacher doesn’t know who they are, why would they trust the teacher’s benevolence if it comes to conflicts? Students therefore sometimes strategically make sure the teacher knows who they are.
• Navigating expectations of teachers: Students report that they expect high AI literacy from teachers, reflected both in conversations about AI but also in changed assignments. They also expect transparency from teachers about their own AI use (but at the same time don’t seem to realise that their own use of AI detection services on teaching materials isn’t 100% foolproof either).

These findings have a bunch of implications for teachers:

• Transparency: Teachers need to let students know about their own GenAI practices, as well as how, for example, disclosure of using GenAI is factored into grading
• Skill: Teachers need to develop and demonstrate GenAI skill
• AI detection: Students need to know the mechanisms both so they can be more informed in what they can actually use without fear, and to also feel more empowered to defend themselves in case of wrong accusations
• Enforcement: The focus should be on discussing and negotiating academic integrity, not on catching out students and punishing them (see also really nice article on academic integrity enforcement pyramid)
• Policies: Should be co-created and continuously re-negotiated in the evolving landscape

This is so helpful, and the article won the very well-deserved Teaching in Higher Education Editors’ Choice Award 2025! It makes me want to talk to students and teachers together about GenAI in their specific contexts, their practices, the university’s framework, etc. So much fun work to be done here!

*In a nice post, “13 Ways to Detect AI Writing Without Technology”, Erik Ofgang shares a list of indications of potential GenAI use, like a student suddenly over-delivering in terms of text length beyond what is asked for while missing the point, a whole bunch of lists (which, to be fair, I do a lot myself), way too many clichés, pattern repetition, and made-up citations. This is an interesting list to talk about with students early on!

Featured image, and image below, from an after work dip a couple of days ago, when I had just started reading this article.

Luo (Jess), J. (2024). How does GenAI affect trust in teacher-student relationships? Insights from students’ assessment experiences. Teaching in Higher Education, 1–16. https://doi.org/10.1080/13562517.2024.2341005