Panelists Discuss AI in K-12 Education at Harvard Ed School Webinar

Education experts discussed how generative artificial intelligence can aid teachers and students in K-12 education at a Harvard Graduate School of Education webinar Wednesday.

The webinar was the latest installment in HGSE’s “Education Now” webinar series, which tracks the effects of the Covid-19 pandemic on American education, with an emphasis on digital educational tools.

Wednesday’s panel was moderated by HGSE lecturer Uche B. Amaechi ’99.

The panelists began by discussing recent reactions to generative AI among educators, which the panelists agreed have focused on the potential for plagiarism and school policies on student work.

Panelist Dora Demszky, an assistant professor in education data science at the Stanford Graduate School of Education, said she aims to make the dialogue around AI more “teacher-facing.”

“Human teachers are pivotal to so many aspects of learning like motivation, like role modeling, et cetera. That’s not something that AI could ever replace,” Demszky said, adding that she believes generative AI is not adequately able to consider the perspectives of minority students or students with disabilities.

Rather than replace teachers, Demszky said she hopes AI will improve the process of professional development for educators. In collaboration with HGSE professor Heather C. Hill and University of Maryland professor Jing Liu, Demszky is developing an AI tool that analyzes teacher performance through audio recordings.

“Currently, teachers receive somewhat infrequent feedback or coaching on their practices due to the resource intensity of such professional learning,” Demszky added. “AI could both help with the frequency and the scalability of all these professional learning programs, but also provide a new perspective, a more private way to get feedback.”

Panelist Lakshya Jain, a senior at King Philip Regional High School in Wrentham, Massachusetts, said he has observed cheating in school using AI, particularly for history or English homework assignments — but added that generative AI should not necessarily be blamed for this.

“They would’ve cheated anyway without ChatGPT. You have all these resources online, so it’s not that ChatGPT is making cheating worse, it’s just making it easier to cheat,” Jain said.

“It also begs the question of, ‘What is the point of an assignment like that?’” Jain added. “If you have projects, project-based assignments, or graded discussions where there’s a lot more involved, I think students are a lot more engaged and they learn a lot more from that.”

“Most of the students that can make ChatGPT write a good essay could write it themselves, maybe in the same amount of time,” Jain said.

Jain said he uses ChatGPT to assist his writing process.

“I put into the prompt, ‘Here’s my article, act as the average American. What do you think about this topic, what are you getting from this article?’” Jain said. “From there I can learn and edit my articles based on ChatGPT’s output to make my articles more informative.”

According to panelist Houman D. Harouni, an HGSE lecturer, generative AI presents a unique opportunity to teach students how to experiment with powerful tools.

“I think what’s happening right now is that something that we might have known in education for a long time, and have never really been able to face, is finally staring at us,” Harouni said.

“We have been focusing way too much on student learning and student outcome,” Harouni added. “Maybe the real criteria is not learning. It’s not even work and performance. Maybe it is agency.”

—Staff writer Azusa M. Lippit can be reached at azusa.lippit@thecrimson.com. Follow her on Twitter @azusalippit.