When novelty fades: Rice responds to ChatGPT with mixed reviews

By Bonnie Zhao     3/7/23 9:39pm

ChatGPT, an artificial intelligence chatbot, has raised concerns in academic institutions over its implications for plagiarism. Rice University’s students, faculty and administrators alike respond to whether ChatGPT has a place in higher education.

Dean of Undergraduates Bridget Gorman said that the Rice administration is in no hurry to change its academic integrity policies to include AI tools like ChatGPT, despite some peer universities such as Washington University in St. Louis amending their policies in response to ChatGPT.

“I don’t really view [ChatGPT] any differently than any other evolution over time as technology has changed … This is just another notch along the way,” Gorman said “It’s still very new … If you’re going to stand up and change a policy, you need to be really confident and understand why and  what you hope to achieve with that change … because these have real implications for students.”

---

---

Gorman said the ultimate judges of ChatGPT’s appropriate use in the university are the faculty senate, the individual faculty members and then the students. 

“At the end of the day, [professors] can still call their class in and sit everybody down and say, ‘Here’s a blue book, get out a pen. You have 15 minutes to answer this question.’” Gorman said. “There’s a lot of different ways that [faculty] could try to think about interacting with [ChatGPT].”

Rice professors have reacted to ChatGPT in various ways. English professor Amanda Johnson, for instance, added a syllabus policy regarding papers that read as if they are written by ChatGPT.

“My policy states that such papers must be completely redone so that the new document presents an original argument. The student must also support this original argument with specific examples, detailed analysis and verifiable citations,” Johnson wrote in an email to the Thresher.

Johnson said when ChatGPT launched, she immediately saw how students unsure about their writing skills might be tempted to use it to generate a paper. However, she actively dissuades her students from using ChatGPT by pointing out its many  drawbacks.

“ChatGPT’s ‘essays’ offer arguments that, at best, are so broad and basic that they could apply to almost anything,” Johnson said. “For example, ChatGPT told me that the 2019 film ‘Knives Out’ ‘is known for its wit, strong performances and a clever plot that keeps the audience guessing until the end,’ but the same could be said of ‘Reservoir Dogs’, which is a very different film that came out in 1992. Any student paper that relies on such statements, then, will ultimately fail to convey a clear, original, consequential argument.”

Moreover, Johnson said when ChatGPT is prompted to make specific or elaborate statements about a topic, it is prone to making statements that are, quite frankly, wrong. 

“When I asked it to summarize the Edgar Allan Poe story ‘The Black Cat,’ the summary it produced contained a factual error in almost every sentence,” Johnson said. “Because ChatGPT cites zero sources for its claims  and because new technologies often carry with them a deceptive sense of authority, I am most urgently concerned with students asking it research questions and becoming the victims of misinformation.”

BioSciences professor Jon Flynn, on the other hand, has allowed his BIOS 442 class to utilize ChatGPT for their most recent exam.

“For this exam, AI assistance may be used as much as you like … However, please keep in mind that experimental design often has a delicate logic to it, and it is something that Large Language Models (LLMs) like ChatGPT routinely have trouble with … Without knowledge in the field to provide an anchor, they can rapidly lead you astray,” Flynn’s assignment reads.

Celeste Wang, a student of this class, said this specific exam is composed of five free response questions that requires them to critically think about the research papers they’ve read, methods they’ve learned and how to apply them. 

“[Flynn] literally told us, ‘You can use ChatGPT. I ran my test through it, you’ll get a B if you do it.’ And then he told us about how ChatGPT … makes you sound good, but doesn’t really answer the questions. But he said … it’s a good starting point,” Wang, a Lovett College junior, said. “I think that’s really helpful because I at least know what I can work with, that there is a beginning point, rather than scouring the entire internet.”

Pedro Ribeiro, the external vice chair of the Honor Council and a co-chair of the council’s AI task force created early this semester, said their official stance on ChatGPT is students shouldn’t use ChatGPT without crediting it.

“Basically, the Honor Council treats ChatGPT as any other source: if you use any of its ideas, you must cite it. As for using it to add grammar or style, that is up to the discretion of the professor,” Riberiro said. “We want students to be submitting their own ideas and their own contributions, both because we want people to learn, and because we want to be a respected institution. But I think [ChatGPT] definitely has good uses that are well within academic integrity.”

Ribeiro said that the AI task force has sent a statement to all faculty, which indicates that utilizing AI software to generate ideas without crediting it should be considered plagiarism, and encouraged them to put it in their syllabi.

“If something is bad, you need to put it on the assignment directions, because otherwise [Honor Council] can’t adjudicate it,” Ribeiro said. “If we’re going to trust students to follow the Honor Code, we need to be clear on what the code is. Because for some classes, maybe it is allowed to use ChatGPT. And something we’ve been really emphasizing to professors is to be clear — as clear as possible.”

Ribeiro said the AI task force has been testing some plagiarism detection softwares but need more data to make a recommendation to the Honor council. 

Elysia Wu, a Lovett senior majoring in Neuroscience and English, said  she primarily uses ChatGPT as a tool for literature review.

“I feel like the benefit of ChatGPT is that it can source a lot of information at once, and that can help a lot when you’re parsing through a lot of information and attempting to write a paper that needs to be informed by a lot of sources,” Wu said. 

Wu said that despite ChatGPT having constraints and limitations, it is overall still a benefit to people who are looking to learn and use the internet to be more efficient and productive in their work.

“I think asking students to memorize quizlets and stuff like that … has been long overdue anyways. I think it’s a real reality that whenever you’re in the workplace, you can probably just Google something,” Wu said. 

Clayton Ramsey, a Baker College senior studying computer science and electrical engineering, said he believes ChatGPT is often “confidently wrong.”

“It’s very easy to get a good-looking result that means absolutely nothing …  I spend as much time verifying [the codes it produces] as I would have to just write it from scratch myself,” Ramsey said. 

Ramsey said using ChatGPT on assignments should be handled the same way as getting help from a friend.

“If I talked to a friend about my essay, and they had a good idea I wanted to go explore, I think that’d be fun,” Ramsey said. “But if I talk to my friend about an essay, and I just lifted their words straight out of their mouth onto my paper, that would be dishonest.”

Some Rice students have used ChatGPT during exams and quizzes. An anonymous senior said they seeked ChatGPT’s help during a take-home exam to check their answers.

“I had this problem that asked me to make a diagram, and I pasted the prompt into ChatGPT. I compared my response to ChatGPT’s, and I think [ChatGPT is] wrong on some tiny details, so I didn’t end up using its answers … I think ChatGPT is not that helpful during exams because it’s not guaranteed to be 100% correct,” the source says. “Also, it crashed, and I had to restart it before it gave me a valid response.”

Another anonymous senior said they utilize ChatGPT for discussion posts and lab assignments. 

“It helps me find resources like scientific articles to quote and helps me with the smoothness of my writing. I have been able to make higher grades on my assignments because of [ChatGPT],” the source says.

Gorman said using ChatGPT in a way that violates Rice’s honor code is a high stakes game that’s low in return.

“If you choose to engage in behavior that’s found to be academically dishonest, the consequences are potentially quite severe,” Gorman said. “Don’t do it. Not worth it.” 

---