Keep academic advising human only

This semester, Harvard implemented a custom-built artificial intelligence chatbot called Harvard Student Compass to aid students in navigating academic requirements and course planning. Built with OpenAI’s ChatGPT Edu software, the tool aims to consolidate the university’s undergraduate advising information all in one place, so students aren’t left searching sprawling websites to figure out what courses they need to take.

The supposed aim of this software is noble. As we know at Princeton, it can be difficult to wrap one’s head around a sea of distribution requirements, major requirements, and class prerequisites all to plan a semester’s courseload, much less a holistic path through the University. 

But implementing a chatbot ignores the root issue — insufficiently tailored advising — and the tool’s ineffectiveness has already been demonstrated by early student reviews. Rather than following Harvard down the AI rabbit hole, Princeton and other universities should consolidate academic information and advising resources in one place and strengthen existing advising programs so that advisers are truly equipped to help students make academic choices.

The first problem with implementing such a tool is the inaccurate nature of chatbots themselves. Most people know that chatbots are prone to hallucinations. But what isn’t mentioned enough is how these fundamental inaccuracies persist at advanced levels due to inherent limitations in the design of generative AI. As experts like James Zou have noted, chatbots are language-predicting tools at their core, built on imperfect training data. They were never designed to tell fact from fiction, and they’re only getting more confident at presenting lies as they advance. That makes them a particularly bad choice for academic advising, which relies on synthesizing exceedingly specific academic requirements that are subject to change. 

Reviews of the existing chatbot at Harvard already show the weaknesses of such a tool. A professor in the school’s statistics department told the Crimson that the tool faltered — and often delivered false information — when prompted with specific curriculum questions likely to come up in meetings. Students affirmed this experience. The chatbot was reportedly somewhat helpful in answering basic questions or finding specific courses that fit a requirement. But this small level of utility is not worth it if the chatbot might mislead students entirely when the questions get tough.

Tools like these also miss the true purpose of academic advising: to deal with those tough questions. While confusing websites are one reason students might need more advising, its true purpose is to give students thoughtful, big-picture advice about their plans. That might involve synthesizing multiple possible major paths or course schedules, but it more so involves thinking holistically about each student’s path. 

Chatbots might be good at emulating this type of mentorship, but they certainly don’t replace the perspective of a true academic professional. As we know from the perils of AI therapy, chatbots are likely to validate the user and are incapable of true critical thinking. Even if the supposed aim of tools like Student Compass is not to provide this type of support, it’s inevitable that students with lackluster human advising might seek it out. Indeed, multiple past op-eds by students at other universities lament the lack of personalized attention from advisers. And Harvard itself, amid 165 layoffs in the Faculty of Arts and Sciences, is beginning to decenter individual, personalized advising in favor of group sessions as it implements Student Compass.

Some might argue that as imperfect as tools like these might be, they at least represent a start toward making academic planning easier for students. But the very existence of Student Compass ignores this root problem of advising at schools like Harvard and Princeton, instead implementing a band-aid solution. In doing so, it only makes true progress more difficult.

And human advising is more important than ever in an academic landscape where AI is beginning to dominate. Professors at myriad universities across the country are struggling to combat AI in the classroom. Industry leaders continue to parrot anti-intellectual myths about how the technology might render traditional education obsolete. And beyond Harvard, some universities are starting to roll out sponsored AI tools that make even less sense than Student Compass. No, not all uses of AI are the same, and Harvard isn’t endorsing the use of AI for academic work by implementing Student Compass. But when generative AI threatens academia in so many ways, both structurally and symbolically, it’s nonsensical for colleges to encourage its uptake through a tool that doesn’t really work in the first place. 

 Resources that could be put toward investments like Student Compass can — and should — instead be put toward improving human advising and clarifying pedagogical policy. At Princeton, this starts with intentional, active advising training for professors that emphasizes the importance of holistic mentorship in choosing an academic path. First-years should be matched with advisers who have actively studied the disciplines in their general field of interest, and they should have the choice to change advisers to a different field at any moment. It’s worth also creating a formalized program to pair students with upperclassmen of their prospective major(s) to help sort through distribution requirements and potential paths to their degree. Finally, Princeton should consider incorporating all major and minor requirements across degrees into one visualized website page so that students can easily compare different paths.

With their enormous endowments, universities like Princeton and Harvard have the resources to make academic choices easier for students. They also have the responsibility, as top universities, to mount a meaningful resistance to AI’s onslaught on higher education. As my colleague Charlie Yale noted in July, schools like UChicago Law are already forging a path to combatting AI in the classroom itself. For both symbolic and practical reasons, we must also resist the implementation of tools like Student Compass and push universities to offer personalized, human academic advising that matches the quality of their education. 

Shane McCauley is a junior from Boston who serves as an associate Opinion editor. He can be reached at smccauley[at]dailyprincetonian.com.