Currently reading “Examining Student and Teacher Perspectives on Undisclosed Use of Generative AI in Academic Work” by Adnin et al. (2025)

Yesterday I wrote about how disclosing the use of GenAI erodes trust, but that it is even worse to get caught using GenAI without having disclosed it, So how do students handle this? In their study “Examining Student and Teacher Perspectives on Undisclosed Use of Generative AI in Academic Work“, Adnin et al. (2025) investigate when, why and how college students hide that they have used GenAI, and how teachers find out about anyway and how they then react to it.

They find that GenAI use is very common, and for many tasks, and that students are strategic about how they use it and if and how they disclose it: “students tried to strike a delicate balance between collaboration with peers and safeguarding their academic “tricks” to maximize personal benefits while minimizing risks associated with disclosure of GenAI use“. The most common reason (54%) that students agree to as reason for non-disclosure is “I feel others will judge me if I disclose it“, which is an accurate threat assessment according to the study I read yesterday. Other common ways non-disclosed GenAI use is justified are “I feel that using GenAI chatbots is similar to using other technologies like Grammarly” (66%), “I feel that instructors are also using it to create assignments” (57%), and “I feel that it is not harmful to anyone” (69%). Additionally, Adnin et al. (2025) find that “[t]he lack of a standardized approach to GenAI across various courses results in students having to navigate a patchwork of different rules and attitudes of instructors, which affects their willingness to disclose their use of GenAI.”

However, students don’t use GenAI for everything. They evaluate, for example, the importance of intellectual contributions to tasks; if something feels like more of a routine task, like checking grammar or developing a standard demographic questionnaire, then GenAI use feels ok and disclosure doesn’t feel necessary. In contrast, for something where own intellectual contribution was expected, it would not feel ok to use GenAI, or at least not only. Then, students would use strategies to hide the use of GenAI, like editing outputs to match own style before submitting, or adding comments to generated code.

Adnin et al. (2025) use cognitive dissonance, the “psychological discomfort experienced by an individual when there exists an inconsistency between their beliefs and their actions“, to explain what is going on. Using GenAI despite knowing that they should be doing the work themselves causes discomfort, and “[t]his discomfort prompts a recovery process to reduce dissonance—either by changing actions to align with beliefs or changing beliefs to align with actions—by changing one of the conflicting elements“. Such coping mechanisms where actions are changed to match beliefs are, for example, editing output repeatedly before submission, or only using GenAI for tasks that don’t seem to require a lot of intellectual input, as well as disclosing usage very selectively. On the other hand, sometimes beliefs are changed to match actions, e.g. claiming that GenAI is just a tool similar to google or Grammarly; claiming that everybody else is using it too; or that tools will be available, and usage accepted, later in their professional life; claiming that it is the better resource than what is provided to them, and claiming that they need to use it because the pressure is so high. I write “claiming”, because “the resolutions that students arrived at—both in terms of adjusting their actions and their beliefs—were often misguided“.

Teachers, on the other hand, (believe they) recognise pattern and use GenAI detection tools. And they generally understand why students use GenAI: because students are overwhelmed, under time pressure, or perceive tasks as busywork. Therefore, “teachers developed a plethora of strategies to manage students’ undisclosed GenAI use, ranging from clearly outlining rules in the syllabi to developing a trust-oriented learning environment.”

Adnin et al. (2025) end with the following recommendations to teachers:

• redesign tasks to make them more meaningful, and explain how they contribute to learning
• debunk myths to “reform[…] students’ misguided behaviours” (see coping strategies for cognitive dissonance above)
• be transparent about expected use (both what and why)
• engage in reflective conversations

Sounds reasonable!

Adnin, R., Pandkar, A., Yao, B., Wang, D., & Das, M. (2025, April). Examining Student and Teacher Perspectives on Undisclosed Use of Generative AI in Academic Work. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (pp. 1-17).

And now it’s time for some wave watching pics from a recent dip! If that was my boat, I would sit in it and read almost every day.

When I am not dipping, that is.

I love how you can juuuust look into the water here at steeper angles to see whatever stuff is growing in there, before the blue of the reflected sky takes over!

Waves are just the best, make me happy without fail.