California Professors Grapple with AI's Impact on Education

Artificial intelligence has significantly altered the academic landscape, prompting some of California's leading university professors to revert to traditional methods like paper and pencil exams. Educators at institutions including UCLA, USC, UC Berkeley, and Stanford are exploring various approaches to manage the integration of AI in their courses.

Some professors are opting to eliminate technology from assessments entirely, focusing on in-class written work as the primary measure of student understanding. Morgan Polikoff, a professor of education at USC, described the situation as "The technology has been shoved down all our throats and we’re all just dealing with the fallout." Polikoff has planned to increase the number of sit-down, paper-and-pencil tests for undergraduate assignments, a move similar to UC Berkeley math professor Richard Borcherds, who has stopped grading homework for correctness.

Polikoff recounted an instance where a freshman submitted an AI-generated paper despite a ban, an easily detectable infraction that resulted in the student being required to rewrite the assignment from scratch. He emphasized the growing difficulty in assessing genuine student knowledge, stating, "I think in general we will need more standardized and AI-proof measures of everything, both in college admissions and once students are enrolled at university." He added that submitting AI-generated work is "a waste of my time and theirs."

USC permits individual instructors to establish their own AI policies. The university also provides a ChatGPT Edu subscription for approximately $3.1 million annually, covering 80,000 users. Polikoff noted this cost is relatively minor compared to other university expenditures.

At UCLA, Thomas Maierhofer, a lecturer in statistics and data science, has adopted a hybrid model. His exams are conducted in person, on paper, without electronic devices. Homework and lab assignments outside the classroom are graded for completion. Maierhofer views AI as a potential educational aid and is developing a custom chatbot named ALISSTAIR to support student learning. He believes that in-person exams are key to academic integrity, stating, "As long as the exams are in-person and on-paper students are only hurting themselves if they use AI to cheat instead of using it to enhance their own learning."

Richard Borcherds at UC Berkeley has shifted grading primarily to in-person exams without electronic devices, citing that online exams have become less reliable due to AI use. He observed a significant increase in average scores on online exams, making it difficult to discern student performance versus AI assistance.

Stanford Law School professor Julian Nyarko holds a more optimistic view of AI's potential in education. He suggested that AI could positively impact learning, especially when contrasted with less effective human instructors. Nyarko's research involved law professors preferring AI-generated answers to student questions 75% of the time over their own colleagues' responses, though this study focused on answer quality rather than long-term learning outcomes. He cautioned against irresponsible use of AI, such as having it complete assignments directly.

Nyarko noted that many professors who previously assigned lengthy take-home exams now feel unable to do so due to concerns about students using AI. He views restricting student access to AI as impractical, given its widespread availability. "AI will require significant change in how we do things," Nyarko said. "We’re basically moving from a world where we kind of thought we knew how education works reasonably well, to a world where we don’t know how to best educate students."

Stanford is offering students free access to ChatGPT, Gemini, and Claude through a pilot program until August 2027, allowing AI use for learning but prohibiting submission of AI-generated work as original.

The debate over AI in education has also entered the political arena in California. Sonja Shaw, the Republican candidate for State Superintendent of Public Instruction, argued that over-reliance on AI hinders the development of critical skills. She advocates for a return to academic rigor and objective testing, stating, "Objective testing remains the best means to measure student skills in an environment where they can’t simply rely on ChatGPT to do the work for them."