Expert system is typically promoted as a way to make teachers more reliable by helping them write lesson strategies, generate class materials and provide feedback to trainees in seconds. But one of the very first randomized trials checking AI in genuine class found that it can also undermine learning. Students whose instructors were given access to an AI teaching assistant feltless motivated to find out.
The damage was particularly noticable among trainees whose instructors were already weaker instructors, as determined by their efficiency before the experiment started. Their students likewise scored lower on standardized last examinations, the researchers found.
“Teachers, similar to students or coders, might be using AI as a crutch,” stated Alp Sungu, lead author of the research study and an assistant professor at the Wharton School at the University of Pennsylvania. “Rather of doing the real work, they’re utilizing AI to entrust the task, and that lowers the quality of their teaching.”
A draft of the research study, “Generative AI Can Damage Mentor,” was released online in June and has not yet been released in a peer-reviewed journal. It echoes Sungu’s extensively gone over 2024 research on how trainees’ usage of AI is hurting knowing.
“Trainees use AI as an answer device, not as a tool for knowing, and therefore it harms learning,” stated Sungu. “Here, I think instructors are possibly utilizing AI as a product creating device for homework, lecture notes, lesson strategies, curriculum. Rather of enhancing their own output, they’re using AI as a replacement with really minimal interaction, and for that reason the quality of output is not good enough.”
Related: Faster solutions, lower test scores: How AI is deteriorating math skills
Sungu’s experiment, carried out with fellow University of Pennsylvania scientists, consisting of academic psychologist Angela Duckworth, followed 193 instructors and more than 2,800 middle and high school trainees in an independent school chain in Turkey during the spring of 2025.
Teachers were arbitrarily appointed either to get access to a ChatGPT-based mentor assistant personalized to Turkey’s nationwide curriculum or to continue teaching as normal. Over 10 weeks, instructors mostly used the tool to create lecture notes, tasks and examinations.
Students whose instructors had access to the AI tool ranked their classes as less pleasurable, less intriguing and lesser than students in the control group. The decline in intrinsic motivation was modest, however bigger among trainees of those teachers who had already been heavier AI users before the experiment began.
Typical academic accomplishment did not alter general. But among teachers whose students had lower marks before the experiment– a proxy for lower-performing teachers– trainee accomplishment and self-confidence both declined. Academic achievement was determined through externally administered standardized tests, dismissing the possibility that these instructors had various grading requirements.
The research study can not describe precisely why mentor quality deteriorated. Researchers did not observe classrooms or evaluate the AI-generated materials teachers used. However Sungu suspects that instructors might have been giving up among their most effective tools.
“When you begin using AI-generated material, you’re losing your personal voice,” stated Sungu. “It might be technically sufficient, however it doesn’t really carry your own style. If whatever is extremely consistent, it simply ends up being a bit more dull.”
One possible description for the weaker scholastic performance among trainees of low-performing teachers, Sungu stated, is that stronger instructors deal with AI output as a first draft, modifying and adapting it to their class. Weaker teachers, he thinks, might be more likely to utilize AI-generated product as is.
Related: AI provides more appreciation, less criticism to Black trainees
This study is not a clean contrast in between teaching with and without AI. Teachers in the control group were complimentary to use other AI tools, making this a contrast between access to a customized AI assistant and whatever teachers chose to do by themselves. If anything, Sungu said, these findings may be understating the threats of teachers relying heavily on AI-generated materials.
Still, Sungu cautions that it would be a mistaketo conclude that “AI is awful and will destroy education.” He sees a different lesson: Access to AI technology alone does not improve teaching.
The challenge is to help teachers utilize AI in manner ins which protect human judgment and imagination. That will need instructor training programs, guardrails and much better user interfaces.
“As of today, how teachers are using it organically, there is something to be worried about,” he stated.
Sungu says he personally uses AI in his university mentor to produce interactive video games and surveys that would otherwise take too long to construct. “When I first get the output, it simply looks excellent,” he said. “And after that, if I don’t immerse myself in it, the examples, the numbers don’t make sense. I end up spending an equal quantity of time to enhance the output or calibrate it to my class.”
“It’s not a convenience,” he said.
Contact staffauthor Jill Barshay at 212-678-3595, jillbarshay.35 on Signal, or [email protected].
This story about AI in teaching was produced by The Hechinger Report, a nonprofit, independent news organization that covers education. Register for Proof Pointsand other Hechinger newsletters.
Was this story valuable? Leave a tip to support your education reporters.
The Hechinger Report is a nonprofit newsroom powered by reader assistance
Republish our short articles totally free, online or in print, under an Imaginative Commons license.