Various Cornell classes across departments have begun integrating artificial intelligence to supplement students’ learning and improve efficiency.
Cornell recently appointed Prof. Thorsten Joachims, computer science and information science, to the new position of vice provost for artificial intelligence strategy on Jan. 1.
In a statement to The Sun, Joachims wrote that when he taught CS 3780: “Intro to Machine Learning,” he allowed the use of generative AI for projects. Instead of students spending many hours “stumping the TAs” while trying to debug challenging code, students were able to ask questions relevant to the course content.
“I saw fewer struggles, and they could engage with the TAs on the interesting machine learning questions — not why their code was throwing some mysterious error,” wrote Joachims.
Vice Provost for Academic Innovation Prof. Steven Jackson, information science, wrote that the use of chatbot tools has allowed students to receive immediate answers for simple questions, mitigating the need for them to rely on office hours for minor questions.
“We expect these experiments [of classes using chatbot tools] to continue and potentially expand as we see evidence that they strengthen student learning and engagement,” Jackson wrote in a statement to The Sun.
While the use of AI for supplementing instruction increases, the Cornell GenAI Education Working Group, including faculty, staff and students is where questions regarding AI in the classroom are discussed. This group is part of the University’s AI Advisory Council. Their goal is to ensure that students receive quality education, feedback and grading to their students.
These goals are implemented through conducting surveys on GenAI, campus-wide programs regarding Critical AI Literacy and working with Cornell Information Technologies and researchers to assess new software, according to its website.
One newly incorporated AI is HiTA, which has been utilized by departments including language, information science and computer science programs, and across schools including the S.C. Johnson College of Business and Weill Cornell Medicine, according to Prof. Rene Kizilcec, information science.
HiTA, which provides academic assistance for higher education through teaching and learning according to its website, operates differently from generative AI platforms like ChatGPT, Microsoft Copilot and Google Gemini.
“[Generative AI] is designed to give you answers fast, to solve your problems fast, and that is not desirable in education, where learning is a process that requires learners to engage in effortful deliberative, intentional processes,” Kizilcec said in an interview with The Sun.
Unlike these generative AI chatbots, HiTA does not directly solve students’ homework problems. The goal of the site is to help students with concepts by providing them with hints, according to Kizilcec.
Cornell classes typically use Canvas as their site for course organization, announcements, homework and exams. In classes like INFO 4100: “Learning Analytics,” which is taught by Kizilcec, HiTA is used instead, offering course assistance and feedback to students.
In INFO 4100: “Learning Analytics,” course materials are released through HiTA, where the program grades homework assignments alongside “comprehensive verification by TAs,” according to the course's syllabus. Students are able to ask a HiTA chatbot for help with homework assignments.
HiTA is specifically trained to be knowledgeable about the course it is used in so that it can assist students.
“HiTA is grounded in the materials of the course and that is critical because every course is slightly different,” Kizilcec said. “You don’t want to confuse students with terminology and concepts that are not actually in the course.”
The software also allows the instructors to monitor the AI. Kizilcec said he has oversight over HiTA with the ability to review the chatbot’s responses to ensure that its answers are not too technical or too vague.
Kizilcec said that he would want HiTA to respond to students in a way that he would respond “if [he] had infinite time to respond to every question at every odd hour of the day.”
Cameron Pien ’27, a student in INFO 4100, stated that HiTA improves efficiency within the class because students are able to be helped at all times of day. In previous classes that did not use AI platforms, Pien said that it would take a long time for students to receive help on their code.
“If you had to ask a human TA for help with code, you would have to wait a long time,” said Pein. “It would sometimes take them a while to figure out how your computer interface works specifically.”
In addition to assisting students, Large Language Models are used to summarize the conversations that students have had and determine the main themes that occurred through conversations.
Kizilcec will look through the summary to understand common questions and misconceptions. Based on these results, he “bring[s] it up in the TA meeting as something that [they] should cover in section” if it is relevant, Kizilcec said.
AI assistance has also found its way into non-STEM courses. Another course that has incorporated AI is SPAN 2090: “Intermediate Spanish I”. The class uses Chitter Chatter, which can be used as a conversation partner to help build language fluency, according to its website.
Elly Burlier ’29, a student in SPAN 2090, said that in this class, students can receive extra credit by completing weekly assignments that include having a conversation with the AI. Chitter Chatter talks to the students in Spanish, and once the student’s work is submitted, the AI gives the student feedback on how they can improve.
Burlier expressed her apprehension towards utilizing the extra credit opportunity.
“I struggle with it because I am ethically very against AI,” Burlier said. “I acknowledge [the] environmental impacts and I think human dependency on it is a problem and I don’t love that it is being introduced into education so much.”
At the same time, Burlier wants to be able to benefit from the extra credit opportunities, and the feedback that Chitter Chatter provides.
“I also do find the platform actually pretty helpful,” said Burlier. “It is very nice to be able to practice conversation [in] such a low stakes way because you’re just talking to a computer.”
Pien told The Sun that she has noticed a varied and gradual acceptance of AI from her professors.
“I feel like I’ve noticed a lot more diversity in the way my different professors have laid out their AI policies,” Pien said. “I see a lot more flexibility in the way professors are incorporating it and maybe seeing it [as] less of a threat and more like a societal inevitability.”
Pien explained that while AI in her STEM classes has been implemented smoothly, she would feel more negatively if it was implemented in a humanities course.
“I think if you’re talking about code where it’s pretty straightforward what the right answer is, then [it’s] completely fine,” said Pien. “But, if this was implemented in a humanities class, I would feel more negatively about AI grading my work.” Specifically, she is weary of AI grading student essays as the grading criteria is less straightforward.
Kizilcec acknowledged that there are nuances in when AI should and should not be used in education.
“I would not advocate that everyone just uses AI everywhere,” Kizilcec said. “The thing that is really important is to be very deliberate and intentional with how we integrate AI into the work that we do.”
The increased use of AI in the classroom does not diminish the responsibilities placed on faculty. They are still expected to play an active role in the classroom and be involved with their students.
“Instructors are ultimately responsible for the accuracy and validity of any grades assigned, and we take that responsibility seriously,” Jackson wrote. “We view all of this work as extending and augmenting faculty and TA feedback, not replacing it.”
