Wednesday, Oct. 1, the College of William and Mary announced the pilot launch of ChatGPT Edu. This comes amid the College’s broader push towards artificial intelligence development, as the College announced an AI minor in August and plans to additionally expand its bachelor’s and graduate opportunities in AI.
The rollout of ChatGPT Edu also reveals divisions at the college over the use of AI. For many students and faculty, the development raises academic and environmental concerns.
Chat GPT Edu will provide select faculty with more OpenAI resources — higher message limits than Chat GPT’s free version, data analytics, web browsing and document summarization. This initiative is sponsored by the School of Computing, Data Sciences and Physics, Information Technology, William and Mary Libraries and the Mason School of Business.
In an interview with University Communications, Chief Information Officer Ed Aractingi imagined the launch as an improvement that will provide access to AI through a central and secure platform. Many administrators characterize the initiative as a marker of the College’s leadership in higher education innovation.
“We are creating opportunities to apply the technological power of AI to diverse domains, sparking new ideas, challenging assumptions and integrating a wide range of perspectives,” Dean of CDSP Douglas Schmidt ’84, M.A. ’86 stated in a quote to University Communications.
For physics major Julia Larmee ’28, the program reflects how AI has been integrated into all of her classes and research. She uses it for code-checking, but doesn’t always view it as a driver of innovation.
“It’s a really great tool because in my research lab, we use it to code — my graduate student uses it to debug his code and fix his code,” Larmee said. “But you can’t use it for everything. We have to be aware of how we’re using it, and double-checking its work.”
In the physics department, several of Larmee’s professors have sparked discussions on the bias and ethical uses of AI.
“What sources is the AI actually pulling from? Is it actually getting the right things?” Larmee said. “If you’re feeding the bot specific studies, how are you avoiding that bias?”
A tenet of ChatGPT Edu is that it doesn’t use institutional data from universities to train its models.
“To my knowledge, I haven’t seen one like this, because AI has to be getting the information from somewhere to function,” Larmee said. “I was not aware that a model could work like that.”
Larmee emphasized the need to learn the fundamental concepts before turning to AI.
“In my coding class, it’s really hard to be taught how to code without using AI,” Larmee said. “AI isn’t always right, and so if I’m not learning how to code but AI doesn’t know the right way — that’s an issue.”
Students and faculty also have expressed concerns over the college’s push towards AI development and the lack of clarity surrounding disciplinary policies on AI.
Although the addition of ChatGPT Edu was not sponsored by the College of Arts and Sciences, the move comes amid continued support for the inclusion of AI at the university level, impacting all departments of the College.
Associate professor of religious studies Andrew Tobolowsky expressed his concerns about the impact of AI on the learning environment at the College.
“What you see is declining critical thinking skills, what you see is a declining ability to read and understand information,” Tobolowsky said. “This is not me saying that AI has no uses anywhere, but it has no uses in the humanities classroom.”
Tobolowsky emphasized that the important part of the humanities is being able to synthesize evidence, gather evidence and turn it into an argument.
“The more you involve AI in the process, the less you develop those skills,” Tobolowsky said.
Tobolowsky notes in his syllabus that AI usage is plagiarism, but the enforcement of that is difficult to implement.
“I’ve had students who clearly did use AI. I asked them to rewrite the paper, and they do it again,” Tobolowsky said. “What am I supposed to do?”
Tobolowsky also admitted that the lack of a clear AI policy at the College is challenging when it comes to enacting disciplinary measures, and that the inclusion of ChatGPT Edu could only encourage the dishonest use of AI in completing assignments.
“The university doesn’t have a clear plan for what will constitute demonstrated AI use if you’re referring them to the Honor Council. It doesn’t have a clear plan for the consequences,” Tobolowsky said. “And now the university is providing them with this technology.”
Tobolowsky remarked that much of the College’s push towards AI has highlighted the positive aspects of the technology, not creating a full picture.
“A lot of people have this idea that if you have a technology, you can find a way to make people use it in only the best ways, and it’s just not true.”
For Tobolowsky and other professors, those pushing AI development as increasingly valuable to education are missing the scope of the problem.
“If talking to ChatGPT about the lessons meant that you understood the lessons better, you wouldn’t be hearing these complaints from humanities professors,” Tobolowsky said.
William and Mary Libraries, one of the sponsors of the ChatGPT Edu initiative, have incorporated AI literacy into their research and developed an AI task force.
William and Mary Libraries AI task force hosted “16 AI things in 93 days” – a program meant to familiarize students, faculty, and staff with AI tools, and the ways they should and should not be applied. Chat GPT Edu will augment the tools the library already uses.
“Our overarching framework is information literacy. How do we help students navigate the information ecosystem and help evaluate sources?” said Head of Research Candace Benjes-Small. “AI literacy is almost a perfect fit into that, we already had that way of approaching the topic.”
In research, William and Mary Libraries use tools like SCITE – an AI database solely based on research literature, not the internet at large.
Part of the AI literacy programming is emphasizing the importance of fact-checking and human-centered thinking.
“That was the start of our AI literacy, AI creates these fake citations because it strings words that look like they would answer your question, whether or not they actually lead anywhere,” said Benjes-Small.
Instruction and Research Librarian Camille Andrews stressed the importance of students thinking carefully about the choices they make involving AI.
“Having that literacy of knowing, here’s what AI does well, in some certain circumstances, here’s what they don’t do so well,” said Andrews.
Students also express concerns about the environmental damage AI causes — especially in regards to its water usage. AI requires data centers to run, which in turn use significant amounts of water to self-regulate and function. For many areas across the United States which house data centers, the significant water depletion required to regulate AI has caused a depleted local water supply and lack of access to clean and running water.
Charlotte Muller ’28, a student studying Economics and Environmental Policy, expressed concern over the College’s lack of response towards AI’s impact on the environment.
“Right now, we’re not thinking about the short-term impact that pulling up that water and damaging the environment is having on local communities,” Muller said.
Muller remarked that although exposure to AI could advance education at the College,
its negative environmental impacts are too important to disregard.
“Even if it might be helpful in the long-term to use AI, it is not beneficial in the short-term for these communities or for the environment, especially if we’re not actively mitigating the negative effects that we’re causing,” Muller said.
