Michelle Zhiqing Yan/Sun Staff Photographer

Cornell computer science majors created a generative AI chatbot for Cornell students.

April 29, 2024

Undergrads Create Cornell-Specific AI Tool Designed to Support Students, Help Professors Adapt to AI Landscape 

Print More

Computer science majors Carl Huang ’25 and Mith Patel ’25 believe that generative artificial intelligence is here to stay — even as many professors are restricting the use of AI in their classrooms. 

Inspired by their own experiences using ChatGPT to study for a linear algebra class, they developed a new Cornell-specific generative AI chatbot that provides students with learning tools while helping professors adapt to AI in classrooms.

The generative AI chatbot is designed to provide interactive study material integrated with professors’ lectures and materials. The chatbot can be customized to fit different classes’ needs. 

The development of the AI chatbot comes as the use of generative AI is being examined and contested on college campuses. A July 2023 report by a Cornell University committee on generative artificial intelligence in education urged educators to consider generative AI when developing class objectives.

Patel and Huang emphasized that rather than eliminating or avoiding AI, universities should adapt to these emerging technologies. 

Tailai Ying ‘27, a freshman computer science major, agreed.

“People are going to use it regardless of whether or not it’s allowed, so I think schools should find a way to properly integrate it,” Ying said, adding that molding curricula around AI would make it harder to use the technology dishonestly. 

Patel explained that the chatbot was created as an alternative for professors who have eliminated the use of AI in their classes, noting students would use ChatGPT regardless.

“A lot of professors see the value in ChatGPT, and a lot of professors see the opposite [of that],” Patel said. “We’re [the chatbot is] a middle ground where they can control it, and it uses [their class] materials, and it’s a tool for students to use.”

After initially experimenting with the technical side of the project, Huang and Patel began conducting interviews with faculty members to determine how to best code the chatbot to serve the needs of students and professors in the classroom.

Prof. William Crepet, plant science, expressed his support for AI integration in courses. In his introductory forensic botany class, Crepet encourages his students to practice cross-examinations for their mock trials by using ChatGPT as a mock witness.

“I think the professors that are banning it, [because] they probably saw some students turn in something that [they] just got off [of ChatGPT] without putting any effort into it,” Crepet said. “That’s not good. So we’re going to have to [find] ways to work with [ChatGPT] that are illuminating for the students.”

While ChatGPT sources its answers from the Internet, the chatbot sources its answers directly from professors’ source materials, such as uploaded class notes, asynchronous lecture transcripts and provided slides. This ensures that the chatbot produces more accurate information, allowing professors to have greater control over the role of AI in their classrooms. 

Patel and Huang highlighted the increased access to resources at no cost as a key benefit of the chatbot. For instance, the chatbot aims to provide answers that teaching assistants would otherwise spend time answering, which would reduce long office hour lines.

“We don’t want to replace TAs. It’s more so you can filter all the easy questions [through the chatbot so] you can spend office hours asking deep questions,” Huang said. “We want to act as a tool to assist human TAs and help how courses are run here. 

Ying added that the chatbot might be helpful for people who experience social anxiety, in addition to eliminating office hours traffic.

The chatbot also provides resources for mental health and general student life. In the future, Patel and Huang plan to code a feature in which the chatbot will give students the phone number for the Cornell University Police Department, resident assistants on call and other emergency resources when asked.

Patel explained that they have received a mostly positive reception from the administration and are currently involved in discussions about expanding the chatbot’s usage across the University. 

The chatbot is currently available to students in around 16 classes. Patel and Huang continually receive feedback from professors and students to improve the platform. They are currently addressing concerns about data privacy. 

Eventually, Patel and Huang hope to make the chabot a tool individualized to different interested institutions, but for now, they are focused on integrating the platform into Cornell’s campus. 

“[Adding more] University resources [is] one thing and full integration on campus to all classes,”  Huang said. “And it’ll become more integrated on campus, sort of like a Canvas type of situation where it’d be like second nature to have a tool like that in the class, where the professor can give it to students.”

Avery Wang ’27 is a Sun contributor and can be reached at [email protected].

Update, April 30, 5:21 p.m.: Two quotes referring to the chatbot’s digital accessibility and compliance with security standards by University standards, have been removed to more accurately reflect the current state of the chatbot.

Update, April 30, 1:57 p.m.: This article has been modified to remove the official name and link to the website of the Cornell-specific generative AI chatbot due to legal concerns.