Academic intelligence

From homework-sharing group chats to the classic peek over a classmate’s shoulder, cheating methods have always been present. Recently, however, ChatGPT has emerged as a powerful tool for academic dishonesty. With ChatGPT’s research abilities and text-generating function, students are only a few clicks away from plagiarizing entire sections of AI-written work.

This reliance on AI can have detrimental effects on both students’ understanding of the material and the development of skills.

“There’s the obvious problem of [students] not learning about the topic at all … but it goes deeper than that,” Taylor Ajax ’25 said. “After a project … the teacher will be giving [students] feedback that isn’t actually what they need. Even if they understand the topic really well, they won’t be getting better at writing.”

English teacher John Boylan recognizes the risks of ChatGPT but believes it ultimately won’t pose a huge problem.

“I wasn’t too worried about it because I feel like in the past … plagiarism has been so easy to detect,” Boylan said. “I’ve always believed that my best teaching requires original student thought.”
Boylan believes offering his support and guidance as a teacher throughout the writing process can reduce the likelihood of a student resorting to ChatGPT.

“I think students use tools like ChatGPT, SparkNotes or Googling essays when they feel a certain amount of desperation,” Boylan said. “If I can minimize that as much as possible, I’m not too worried about it.”

According to a survey by Study.com, an online learning platform, 26% of teachers have caught a student cheating using ChatGPT. When it comes to written assignments, plagiarism detectors like Turnitin and OpenAI’s Text Classifier can find cheating. However, there are already methods students can use to bypass these checks, whether it’s manual alterations of AI-generated texts or paraphrasing tools like QuillBot.

“If you put [the text] through QuillBot and … Grammarly or something, and then use synonyms for a bunch of words, all of sudden [teachers] can’t detect AI anymore,” Ajax said.

Built by Princeton University senior Edward Tian, GPTZero is a highly-publicized AI detection tool that has been released to the public in the beta stage. It works by measuring two variables, “perplexity” and “burstiness,” that appear at higher levels in human-produced writing. Still, the tool is not entirely accurate. In a study conducted by a science and technology website called Futurism, GPTZero correctly identified the ChatGPT text in seven out of eight attempts and the human writing six out of eight times.

Without a 100% accurate AI detection tool, schools may have to implement other strategies to keep students from cheating. Patrick Fan, the Tippie Excellence Chair in Business Analytics at the University of Iowa, details one possible approach.

“I think we need to try to educate students,” Fan said. “What is the purpose for [students] to go to high school to learn to write? You want to know how to write properly, how to communicate, how to better position yourself for college studies. And if [students] say, ‘I’m gonna use ChatGPT to finish all my assignments,’ I think that’s gonna be a wrong attitude.”

Hourcade thinks that ChatGPT can be helpful during brainstorming.

“It can be a really useful tool to partner with,” Hourcade said. “And I think its greatest potential [in] education is if you use it as a tool to give you ideas … to help you get started with something.”
Ajax, however, believes many students won’t be convinced to limit their ChatGPT usage.

“If [teachers] try to say, ‘Just use it for an outline,’ you don’t know if students are going to follow that,” Ajax said. “If we could get a version of [ChatGPT] that will just answer your questions about outlining and brainstorming … it would be one of the best learning tools they could [make] for students.”

Various people and organizations are looking to expand on AI’s educational benefits.

Along with three other University of Iowa researchers involved in AI, Demir published a research paper in February designing an AI teaching assistant to use at the University. Using GPT-3 technology, the teaching assistant system would answer course-specific questions and provide services such as summarizing readings and explaining classroom material, much like a study partner.

“It could potentially use student information, like their grades and their exams and work,” Demir said. “[We] don’t need to worry about [students] asking [instructors] the same questions five times in different ways. TAs can spend more quality time with students and provide more advanced support rather than simple basic questions.”

Similar to Demir’s teaching assistant, Khan Academy announced in March that it would pilot “Khanmigo,” an experimental AI guide that mimics one-on-one tutoring using GPT-4 technology.

Taking advantage of AI’s educational benefits requires knowing how to use the tools. According to ICCSD Director of Technology & Innovation Adam Kurth, courses specifically focused on AI are already being designed at the high school level.

“Implementing coursework in this area makes a lot of sense,” Kurth said. “I think that ultimately, one of the things that we need to do is work with teachers to develop new approaches to education in AI in today’s world, that acknowledge AI and even leverage it where appropriate.”

Hourcade believes computer literacy courses should also be a regular part of education.

“I think computer literacy — to understand how the systems work, even just how data gets collected from you, all the time — will be important for people to know,” Hourcade said. “We’re not [all] going to be computer science majors, [but] I would make [computer literacy] a requirement at the high school [level].”

The education system’s expansion of technology-specific courses could address the issue of digital equity, the ability of individuals to fully participate in society, democracy and the economy by having sufficient information and technological ability, as defined by the National Digital Inclusion Alliance.

Hourcade believes government intervention may be necessary to prevent further exacerbation of digital inequity due to the expansion of AI.

“Some folks [who are] better informed and better educated are the ones who are going to likely benefit more from these systems because they’re aware of how they work … Who’s going to get hurt the most is likely going to be people with lower levels of education or socioeconomic status,” Hourcade said. “That’s where you need some government to step in and provide some level of guardrails.”

Ajax sees this digital divide reflected in the classroom.

“It’s not even the students who are struggling in the classes; it’s more so the students who are doing well who just want an easier way of maintaining their grade,” Ajax said.

However, ChatGPT could also help non-native language speakers and people with disabilities. With its language-processing abilities, the software could allow those with learning disabilities or speech or literacy impairments to communicate more effectively by turning almost any input into a more sophisticated output. Additionally, students who receive course content in an unfamiliar language are able to easily translate and simplify the content into their native language, bridging the language barrier.

Fan holds that ChatGPT’s overall potential as a learning tool will naturally close gaps in education.
“[We can] use ChatGPT as a companion tool to help the students in [disadvantaged] situations to learn to improve on their curriculum,” Fan said. “[We can] help guide the students [to do] a better job in the pedagogy in the learning process.”

Boylan intends to continue addressing equity through his process-focused teaching philosophy.

“Someone who [is] tech-savvy in high school would have been able to use [ChatGPT] and gotten a really good grade on [an assignment],” Boylan said. “I think that [it’s] a lot more equitable to create productive struggle for every student than it is to be outcome-focused.”