AI Feedback Could Improve Teaching, According to New Study

A recent study led by Stanford researchers has found that AI could improve the quality of teaching by providing feedback to instructors on their interactions with students in class.

According to the study, employing an automated feedback tool can aid instructors in acknowledging, reinforcing, and expanding their students' contributions, also known as uptake.

Students who received this form of feedback exhibited higher satisfaction with the course and performed better on assignments, on average.

M-Powering Teachers

Dora Demszky, an assistant professor at the Stanford Graduate School of Education (GSE) and the lead author of the study, stated that providing timely and specific feedback can enhance teaching. However, it is neither scalable nor practical for someone to sit in a classroom and give feedback every time.

Therefore, Demszky and her colleagues created a cost-effective alternative tool, M-Powering Teachers, which employs natural language processing (NLP) to examine transcripts of a class session, detect conversational patterns, and deliver reliable, automated feedback.

The study focused on the practice of teachers acknowledging and building on their students' contributions, known as uptake, which is essential for students to feel heard and has been linked to better academic performance.

The tool analyzes the teacher's response to a student's contribution, detecting how well the teacher understood and expanded on the student's idea.

In addition, the tool provides feedback on teachers' questioning techniques, such as asking questions that prompt student responses and the ratio of teacher-to-student talk time.

Testing this AI Tool

To test the tool, the researchers used it during the Spring 2021 session of Code in Place, a free online course that teaches basic programming to students globally.

Volunteer instructors who received feedback from the tool increased their use of uptake and improved their questioning practices.

Additionally, students in the feedback group reported higher levels of satisfaction with the course and showed greater learning gains compared to the control group.

The tool serves as an affordable supplement to traditional classroom evaluations, allowing educators to enhance their teaching without requiring the presence of an instructional coach or other specialists to observe their lessons and create recommendations.

Chris Piech, a co-founder of Code in Place and an assistant professor of computer science education at Stanford, notes that while timely feedback is crucial for students, it is rarely given to teachers.

Piech argues that automated feedback provides teachers with a comfortable way to receive feedback that does not come from their principal, and they can receive it not only after years of practice but also from their first day on the job.

Demszky and her colleagues are currently exploring how the tool could be used in other educational settings, such as one-on-one mentoring programs for high school students.

They hope that the tool's success in improving uptake and questioning practices will encourage other institutions to adopt similar AI-powered solutions to enhance teaching practices.

The study's findings were published in Sage Journals.