Rapidly adapting to AI in education and the library 

Part of the Learning with the Machines feature on how UIC educators and researchers are exploring the impact of large language models 

The ripples of language models go far beyond English classes and writing programs, affecting virtually any class where essays or written exams are part of the syllabus. So, when Erin Stapleton-Corcoran started seeing headlines about ChatGPT in late 2022 and tinkered with the technology, she knew immediately that it would be a paradigm shift for teaching. 

“I guess what was so surprising is that it was so easy to use immediately. It was a very low barrier of entry,” said Stapleton-Corcoran, instructional designer at the Center for the Advancement of Teaching Excellence. “I thought, this pushes the envelope on restructuring and reframing educational practices, the science and art of teaching and learning.” 

Her intuition was right — within days, the center started getting requests from UIC faculty about how to handle the new models. In April, Stapleton-Corcoran and colleagues at the center organized an online discussion about using generative AI in the classroom, attended by more than 50 instructors. The mood on Zoom was divided, she said.  

“There’s concern about the tool, there are some individuals that just don’t like it, don’t want it used by students in their classes,” Stapleton-Corcoran said. “Then there are some that are interested in it, and about how to utilize it in their own teaching workflow, how to redesign assignments.” 

In a blog post and subsequent resource guide co-written with instructional designer Patrick Horton, Stapleton-Corcoran surveyed literature and gathered recommendations around large language models and education. Teachers from all levels can consult the guide for example language limiting or prohibiting use of AI tools in class, ideas for incorporating AI into assignments or designing assignments that discourage its use, and tips for prompting the models effectively and citing their contributions. 

While the rapid response to the fast-evolving field of AI feels dramatic, Stapleton-Corcoran compared it to the disruptions of the COVID-19 pandemic, when long-simmering questions about online education suddenly became inescapable and nimble changes were necessary. And as with COVID-19, the decision-making around these shifts can’t be separated from broader structural concerns. 

“I think there’s a little bit of hesitation about if it’s okay to use the tool: Is it an issue of equity, because some students have access and some students don’t?,” Stapleton-Corcoran said. “We just went through this with COVID, where some students had access to computers and the internet and some students didn’t. I think there are a lot of echoes of that, and faculty are saying ‘let’s carefully consider this technology’ before they fully implement it.” 

That unease is shared by UIC students, according to Anna Kozlowska, assistant professor, reference librarian, and Honors College instructor and fellow. In her Honors College course on media and information literacy, she found a mix of excitement and anxiety from her students about ChatGPT and related tools. 

“Many of them pointed out that it will increase already existing inequality because of the jobs that potentially are at the highest risk of being replaced,” Kozlowska said. “Many of the students are also fearful about their own job prospects and what kind of impact it’s going to have on the job market.” 

At the UIC Library, staff have already seen some requests for papers cited by AI models that don’t exist, and people prompting the “Chat with a Librarian” service on the library website as though it were a language model. But Kozlowska said that there are bigger implications of language models that librarians need to start discussing. 

“In my opinion, we should be talking more about it, and not only in the context of students sending fake citations but on the broader implications,” Kozlowska said. “We need to talk more about how we should be redesigning our instruction to address these tools’ bias, privacy and intellectual property issues.”