STEVE INSKEEP, HOST:
College students are increasingly using AI chatbots to brainstorm ideas or study for quizzes or write their essays. Well, what are professors doing? Lee Gaines reports on data from one AI company with answers.
LEE GAINES, BYLINE: Georgia State University professor Sue Kasun says she used Gemini, Google's AI model, to help her brainstorm assignments for a new course.
SUE KASUN: There were, like, suggestions of offering different choices, like having students generate an image, having students write a poem.
GAINES: The course was about how to integrate culture and identity into language education for immigrant youth.
KASUN: And I - like, these are things that I could maybe think of, but we have limits on our time.
GAINES: Kasun teaches current and future English language educators. She's one of an unknown number of higher education faculty using generative AI models in their work.
DREW BENT: When we looked into the data late last year, we saw that of all the ways people were using Claude, education made up two out of the top four use cases.
GAINES: Drew Bent leads education research at Anthropic. His company found that both professors and students use its AI chatbot, Claude. Their latest data focuses on how professors across the globe use it.
BENT: Developing curricula and study materials was the top use case, but we also saw them using it for academic research.
GAINES: The data comes with caveats. Bent says no humans actually read professors' interactions with Claude. Instead, they used a tool to analyze conversations with Claude among accounts associated with higher education email addresses. That produced 74,000 conversations over an 11-day period earlier this year. Bent says professors frequently automated administrative tasks, like drafting emails and creating budgets. But often, for more complex and creative work like lesson planning...
BENT: The educators and the AI assistant are going back and forth and collaborating on it together.
GAINES: Anthropic published its findings but did not release the full data behind them, including how many professors were in the analysis. But it found that about 7% of the conversations they analyzed were about grading student work.
BENT: And it wasn't the top use case. But it was one of the use cases where, when educators use AI for grading, they often automate a lot of it away, and they have AI do significant parts of the grading.
GAINES: We don't know how these conversations factored into the actual grades students received. But Marc Watkins at the University of Mississippi finds these results alarming. He studies the impact of AI on higher education.
MARC WATKINS: If you're just using this to automate some portion of your life, whether that's writing emails to students, letters of recommendation, grading or providing feedback, I'm really against that for a lot of reasons.
GAINES: Watkins says using AI in this manner could hurt professor-student relationships and undermine the value of higher education.
WATKINS: The sort of nightmare scenario that we might be running into is students using AI to write papers and teachers using AI to grade the same papers. If that's the case, then what's the purpose of education?
GAINES: Kasun, the professor from Georgia State, doesn't believe professors should use AI for grading. She says she's found helpful ways to use AI with her students, but she wishes colleges and universities had more support and guidance on how best to use this new technology.
KASUN: We are here, sort of alone in the forest, fending for ourselves.
GAINES: Drew Bent with Anthropic says companies like his should partner with higher education institutions. But, he cautions...
BENT: Us as a tech company telling educators what to do or what not to do is not the right way.
GAINES: Bent and the professors agree that decisions made now over how to incorporate AI into education will impact students for years to come.
For NPR News, I'm Lee Gaines.
(SOUNDBITE OF ODDISEE'S "BEACH DR.") Transcript provided by NPR, Copyright NPR.
NPR transcripts are created on a rush deadline by an NPR contractor. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.