NOTE: I use the term LLM rather than AI because intelligence, as currently understood, is a biological process — the capacity of a living system to perceive, adapt, and generate meaning from experience and within context. LLMs do none of these things; they predict statistically likely outputs from patterns in text. The term AI obscures that difference more than it illuminates it.
This week I started a guest role at a local university to develop the best use of LLMs for undergraduate students. The opportunity began as a late night debate on the threat of these new tools. The fear was that students would use it to cheat.
We covered 3 possible lines of attack:
- Ban – Outlaw LLMs (good luck with that),
- Regulate – Guide usage of LLMs through rules and procedures. For example, just as students must show their work in math, have them include the exact prompt they used as an addendum on all work. For testing, pivot from rote to aptitude assessment.
- Embrace – Lean into LLMs by incorporating them into classroom instruction as a kind of indefatigable grad student. Deploy them deliberately to guide the students in developing the essentials of critical thinking, writing, research, organizing thought, and creating thesis statements and interesting questions.
The third option was also directed at professors, who tend to be experts in their fields but not always in teaching. Academia is for research and exploration for their own sake. However, the art and science of guiding minds in the process of becoming is a completely distinct discipline. It used to be assumed that this development happened naturally in the rarefied gauntlet of university life. Now society has pivoted to helicopter parenting at all stages, and universities have followed suit, padding the rough edges that once chiseled the mind. The third approach, embrace, uses tech as a tool to liberate professors to focus on their expertise while scaffolding students in skills they need now and for a lifetime.
I drafted the scaffolding: policy, training, and student onboarding. Included were suggestions on how to adjust testing, assign and grade homework, and deliver on the promise of using LLMs to develop critical thinking in the classroom. I created a process to guide students in the prompts for research, organizing data, and writing basic essay formats. It was thorough and included using LLMs to adapt my framework to develop their own living individual voice and writing guide.
On the first day, I arrived ready to present the framework I’d built. Instead, I was told to interview the students. At first, I thought it was just an exercise, so I asked neutral, open-ended questions about their use of LLMs. Their answers shook me.
They were already disillusioned. Relying on LLMs for everything, they felt they had lost the ability to write at all. LLMs produced passable work, but it was generic, void of their soul. They were trapped in a cycle. Already the university was at the mercy of a single voice, not Skynet but something worse: mediocrity. I hadn’t expected their insight or visceral distaste for the shortcut. I had underestimated them, just like the university had. They were not in danger of using LLMs to cheat; they were in danger of being erased.
Undergraduate years are when students must forge their voice, their ability to argue, to defend, and to build worthy concepts. It’s where curiosity and personality take shape, forming the foundations of a rewarding life. Failing now means they will likely never develop these skills later. The professional world assumes you’ve already honed your capacity and are ready to produce more. What was a shortcut temptation at university becomes an expected tool for productivity. If they can’t think conceptually, why hire them at all? LLMs can already do the rest.
Even as I built the approach for professors, I was thinking of the students. I knew the real danger of treating LLMs as a shortcut. Without a deliberate structure for critical thinking, expressive dexterity, and original voice, the students would pay the highest price. What I hadn’t expected was that they already knew and resented it.
The students expressed a helpless ennui, wanting authenticity but finding relying on LLMs a pull too strong to resist.
So, in addition to the work above, I developed a process for the students themselves. I built an adaptive process for creating a self-authored style sheet: a way to prompt both students and LLMs to develop and honor their cognitive identity and voice. Rather than doing the work for them, LLMs could scaffold their growth, with the student interjecting at crucial, strategic points. LLMs would do the heavy lifting, freeing them to climb higher, faster.
It isn’t foolproof, it isn’t meant to be. Caging people or forbidding tools never works. This is for those who take responsibility for their own development. It lets them use LLMs to enhance, not surrender, their soul. Much more remains. The real work is learning how to deploy LLMs for personal growth. The lure of this tool is its radical productivity. But that goal alone is quantity over quality, a dead ocean of mediocrity. The point of any tool is to serve us, not replace us.
