Don’t Sacrifice Critical Thinking for Efficiency

Harvard is sending me mixed signals.

This spring, the Harvard College Writing Program launched a new artificial intelligence training module covering the baselines of how large language models operate in every freshman expository writing course. In my section, where the use of generative AI is strictly prohibited at every stage in the writing process, I am now required to sit through a crash course on the foundations of artificial intelligence.

Students learn the building blocks of AI through a module without any true stance. The module’s designers instead hope to promote students in developing their own positions.

Yet, any course notorious as Expos for building critical thinking skills through writing, is not the environment to have students contemplate the usage of generative AI.

In a recent talk, Harvard College Dean David J. Deming encouraged students to familiarize themselves with the technology. Deming posited the acquisition of AI literacy as a new way to boost productivity, an efficient tool that helps students save time.

Meanwhile, many notable courses have policies forbidding the use of AI in class entirely. While Harvard seems eager to introduce AI as a tool for more efficient learning, the University still needs to make an emphatic delineation between teaching the foundations of AI and its role in the classroom. It’s imperative that students learn to adapt to an ever-changing technological world, but not at the expense of misleading students on what role that exact technology plays in their own education.

Expos is Harvard’s canonical writing class with over 90 percent of undergraduates using it to fulfill their writing requirement. For many freshmen, Expos is their first introduction to college writing. It instills the foundations of building an analytical argument necessary for any future writing they will engage with throughout college, if not life.

More than mere supplementary skills being taught, these are the basics. Learning the essentials of writing is what students should associate with Expos — not the essentials of AI. If the whole premise of Expos is to force students to think critically, suggesting a more efficient alternative directly opposes any skills the College is trying to drill.

The module would like to have its cake and eat it too. On the one hand, it details how critical the process of writing is as a skill sans AI. But at the same time, it shares conflicting messaging conceptualizing AI as a tool to develop new skills. Two different messages are being sent, even though in an Expos class, the former should be explicitly declared to outweigh the latter.

AI, especially positioned as an efficient tool for students, infringes on that pedagogical aim of Expos. The results of a recent MIT Media Lab study suggested that large language models may cause cognitive atrophy as users become overreliant, the literal antithesis to Expos’ philosophy. To insert an LLM 101 in the course is unsettling, if not intellectually damaging.

If Harvard wants to continue on the trajectory of embracing AI, it would need to develop a model of use different than efficiency — and that simply isn’t going to be done in a freshmen writing requirement. Faster doesn’t mean better: Efficiency does not matter in a class if it strips the critical thinking skills college students should be building.

Nevertheless, artificial intelligence is advancing with haste, and it (probably) isn’t going anywhere anytime soon. For Harvard and its students to keep up to date, it makes sense to inform students of the workings of AI. But perhaps the least appropriate setting thinkable for such advice is a class that ideally never yields a project touched by AI.

The logistics of large language models would make a great title for a pre-orientation educational module. Incoming freshmen can (and should) be exposed to this growing field, especially given how ubiquitous it is fast becoming. And at the same time, they can learn that while this domain expands, it is not a replacement for the work of thinking.

Harvard could even enhance the module to propose what using generative AI as a catalyst of critical thinking looks like. AI shouldn’t be the efficient option of shaving off time; it should be an option to grow through trial and error, maybe even the lengthier path of learning.

Harvard has to assert a substantial stance on generative AI. An AI module holds potential, but if Harvard wants to preserve critical thinking, it can’t be through the College’s vestibule of thought.

Olivia R. Nelson ’29, a Crimson Editorial editor, lives in Hurlbut Hall.