Sunday, December 15, 2024

An AI-assisted textbook at UCLA has some academics worried

The University of California, Los Angeles, is offering a medieval literature course next year that will use an AI-generated textbook.

The textbook, developed in partnership with the learning tool company Kudu, was produced from course materials provided by Zrinka Stahuljak, the comparative literature professor at UCLA teaching the class. Students can interact with the textbook and ask it for clarifications and summaries, though it’s programmed to prevent students from using it to write their papers and other assignments.

And as opposed to the nearly $200 students were required to spend on traditional texts—including anthologies and primary-source documents—for previous versions of the course, the AI-generated textbook costs $25.

Stahuljak said it will save students money and give her and her teaching assistants more time to engage students in deeper, more nuanced discussions about the material.

Making Students ‘Brain-Dead’?

But the announcement of the course earlier this week was met with hostility from some in the academic community who worry AI-generated textbooks could lead to compromising not only their livelihoods, but the quality of a college education.

“This is truly bad and makes me wonder if we aren’t participating in creating our own replacements at the expense of, well, everyone who cares about teaching and learning,” Thomas Davis, an associate professor of English at Ohio State University, posted Sunday on Bluesky.

Others characterized the AI-generated textbook as “flat out stupid,” “absolute nonsense” and an idea that takes “the human out of humanities.”

And when the news release about the course was posted on UCLA’s Reddit page on Monday, many users—some of whom alluded to being former and current students at the institution—weren’t thrilled about the idea, either.

“Wow cool. That’s totally worth my tuition money. Not sure I can disagree with ‘College is a scam’ as much as I used to before …” one anonymous user posted on Reddit. “They best offer a huge discount,” said another. “Disgusting,” yet another commenter posted. “So glad we’re using AI to turn people brain dead.”

But not everyone sees the introduction of an AI-generated textbook as a threat.

“I don’t care who or what writes the textbook,” one anonymous Reddit user said amid the deluge of outrage. “This is still taught by a human professor and human TAs.”

That sentiment is what Stahuljak wants critics to understand.

“It allows me to be a professor I’ve never been before but always wanted to be,” she said in an interview with Inside Higher Ed.

Instead of students reading a static textbook before class and waiting until class time for clarification from Stahuljak, the AI-generated textbook she edited has the capability to guide students through the basics of the course material—which has always been a compilation of difficult-to-access primary sources that aren’t contained in any one published work—before they arrive for class. Previously, she said, “I would have to lecture all of the stuff that is in the textbook.”

‘Time That I Never Had’

With the introduction of the AI textbook, “I have time that I’ve never had, which is to work with students to actually read, interpret, think, ask questions, pose different answers, argue different answers and possibilities and hand them over to the TA,” Stahuljak said.

She doesn’t foresee this model putting teaching assistants out of a job, much less professors, but rather elevating their purpose.

“Instead of summarizing my lectures, connecting the lecture to primary sources and going through basic understanding of the primary sources with the students—all of that is done through the textbook and my lectures,” she said. Now, TAs “can actually work with students to create activities that will help students practice reading, writing and debating.”

But limiting the textbook to include only information Stahuljak has fed it—and not drawing on the whole universe of freely available information online that feeds many other large language models—may be “doing a disservice for students in the AI era, who are going to be grappling with the whole of it,” said Hollis Robbins, an English professor and special adviser for humanities diplomacy at the University of Utah focused on AI’s impact on higher education.

“In all humanities classes, the question is always, ‘How do we know what we know?’ which has become the most important question of the AI era,” she said, adding that generative AI is notorious for not properly citing its sources. “What faculty need to do—and administrators need to support—is the larger question of what does higher ed look like when AI has absorbed the knowledge commons, which is going to happen soon.”

Stahuljak said, however, there’s nothing stopping students from using ChatGPT, Meta or any other AI tool on their own time, and that offering a self-contained AI-generated textbook is still an upgrade from traditional textbooks, which are by definition self-contained.

“It’s actually far more interactive than a printed text,” she said. “We’re using a textbook, pedagogical materials and freeing up my time and teaching assistants’ times to work with students to develop critical thinking, reading and interpretive skills to situate something in a context and to be able to work with fiction and history and develop awareness to help them in the real world.”

And for the critics worried that a course with an AI-generated textbook could be duplicated, sold and potentially put professors and teaching assistants out of work, Stahuljak said that’s not what she envisions for the future of AI-infused academia.

“This is a human-run course. It is a human-centered course. And it is assisted by AI,” she said. “If we put this textbook out in the world, we’re doing a kind of community service. But it’s adaptable as well. If I have a particular question that comes up in the classroom that seems to agitate a good number of students, I can go back and work with Kudu to create content based on what I know and what I can provide that students can then consult.”

No ‘Clear Professional Norms’

The launch of Stahuljak’s comparative literature course early next year, following a course in the history department that used an interactive textbook this fall, constitute Kudu’s inaugural interactive textbook projects. Kudu also used AI to help compile a handful of textbook chapters for physics and engineering courses offered at California State Polytechnic University, Pomona, this fall, though those courses still relied heavily on traditional textbooks.

While the history and literature courses are Kudu’s first foray into creating full, customized textbooks, AI has already begun infiltrating the textbook market. In September, Pearson introduced several new AI-powered tools—including personalized study and exam prep, interactive video learning, and AI-powered tutors—to enhance its e-textbooks and study platforms.

Alexander Kusenko, a physics and astronomy professor at UCLA who co-founded Kudu, believes these textbooks have the potential to “make teaching more personal and more tailored to the students’ needs.” He added that a research project underway at UCLA’s Center for Education Innovation and Learning in the Sciences is already showing that AI tools are especially helpful to underrepresented minority students who may “face a social barrier asking questions from their peers or TAs or professors.” By contrast, “there is no social barrier for asking a computer to help.”

Mark Carrigan, a senior lecturer in education at the University of Manchester who wrote the forthcoming book Generative AI for Academics, said in an email to Inside Higher Ed that while the prospect of AI replacing professors isn’t “an immediate threat,” he is concerned “that tasks could be gradually transferred to AI in ways which fundamentally change what it means to be an academic.”

And without developing “clear professional norms around how we use these tools,” Carrigan said, “we risk sleepwalking into an outcome where human scholarship becomes increasingly confined to elite institutions, while the rest of the sector becomes progressively automated in response to financial pressures and institutional incentives.”

Related Articles

Latest Articles