Is AI doing the homework? Teachers are bringing students back to take exams in class

A new study by HIT and the Hebrew University reveals how artificial intelligence is changing assessment methods in schools: many teachers express a preference for exams and in-class work over homework and projects, alongside oral presentations where students are required to prove their knowledge. Almost 90% of the teachers interviewed for the study described situations where they limit the use of artificial intelligence, as it becomes harder to determine who is truly behind the submitted work.

The study, published in the journal Teaching and Teacher Education, was conducted by Dr. Meital Amzaleg and Prof. Gila Kurtz from the Faculty of Instructional Technologies at HIT (Holon Institute of Technology), and Omri Hadar and Prof. Yifat Ben-David Kolikant from the Seymour Fox School of Education at the Hebrew University of Jerusalem. It is based on semi-structured interviews with 24 middle school teachers in Israel.

21 out of 24 interviewees described the need to limit or prevent students' use of generative AI, even among teachers who otherwise integrate the technology into their teaching. When they fear that AI will undermine the purpose of an assignment, teachers prefer in-class work, exams, and oral presentations where the student must explain what they have learned.

The most prominent reason was the desire to ensure that assessments reflect the student's own knowledge. Researchers call this approach "pedagogy of blocking": instead of being sweepingly for or against AI, the teacher examines whether its use in a specific assignment promotes the learning goal or allows the student to bypass the skill they are supposed to perform.

For example, a literature teacher who integrates AI into the learning process said she is considering replacing some home projects with in-class exams. Another teacher of English used the technology extensively for spoken practice and feedback, but limited its use when the goal was to practice independent writing.

The study shows that reluctance to use AI is not necessarily a result of a lack of familiarity with the technology. A mathematics teacher with a master's degree in computer science specializing in AI believed that students' use of generative AI could harm the development of mathematical thinking, as the effort invested in finding a solution is part of the learning process.

In analyzing the interviews, researchers identified four patterns of thinking: developing higher-order thinking, maintaining authority and boundaries, imparting knowledge, and flexibility in adapting to students' needs. However, this did not create clear "for" and "against" camps: the same teacher might integrate the technology into one assignment and limit it in another.

The study also reveals the uncertainty in which teachers operate. 15 out of 24 raised dilemmas regarding adapting teaching to the AI era, 17 feel that the responsibility for decision-making is largely placed on them, and 11 described feeling that they are dealing with the change largely alone.

According to the researchers, training teachers for the AI era cannot be limited to learning new tools. The challenge is to help teachers decide when technology enables better learning and when it performs the thinking or skill the assignment is intended to develop. The researchers emphasize that this is a qualitative study intended to identify patterns of pedagogical thinking.