Earlier this week, one of our multilingual student used AI to write their ESL class to compose their personal narrative essay. The evidence to their teacher could not be much more clear. Their grammar was spotless, the essay structure was advanced, and their vocabulary was far beyond what has been noted throughout the class hitherto that assignment.

I find myself recently constantly grappling with how to get students to utilize AI in class responsibly. The dead-end argument that I always come back to is: Why even bother learning how to write or learn another language when AI can become the middle man in the process? I want to believe that these skills are still essential, I want to believe that students still need to learn it, and I think teachers need to be on the same philosophical page in order to teach them.

Writing is thinking

The practice of writing ensures that you are thinking. Even if a student pays little attention to their word choice, grammar, or style, in order to craft a work that makes sense, they need to ruminate on how abstract ideas can answer some essential question. Once they begin this intense pondering, they sharpen their ideas through these components to more precisely communicate their meaning. If doing it properly, you are piecing together and picking apart your own thought processes bit-by-bit, ensuring they are sound. Like a vulture picking the final scraps off a bone, you methodically hunt down words, sentences, ideas, or sources to tweak so that we can communicate with true lucidity.

I truly (perhaps naively) believe that this skill will endure as an essential element to teach students to perform at a high level, regardless of coming inventions. In the same way that a math teacher requires students to show their work, learning to write cohesively is the most efficient way for students to showcase their cognition, their knowledge and their voice; achieving precisely what you mean, exactly how you mean it.

Technology has always been a gateway for conveyance, eliminating all obstacles that prevent you from converting your personal electrical signals to a readable and transmittable format. From the days of papyrus and soot we have been able to log our knowledge and collect our beliefs that endure through time. By teaching writing, we are presupposing that our thoughts and experiences are valuable, and technology is a vehicle to produce these ideas with ease.

In a way, this is why the mechanical keyboard community has such a large following. They know more than anyone the power of hardware and software as a tool to eliminate friction between your brain and the world. This hearkens back to László Bíró on his invention of the ballpoint pen1:

Well, writing comes from the heart. If we can help the hand to perform the task, what is so wrong with that?

Technology and writing have a deeply intertwined relationship, that is certainly not going away. The problem is when the technology stops acting as a catalyst, and starts to become the reaction.

But that school assignment needs to get done. It's due tomorrow. I am just gonna paste the teacher's prompt and the rubric into ChatGPT, copy and paste, and voila! All done!

This is not to blame them at all; billions of years of evolution always guides us toward the logical path of least resistance.

In the same way that the "Whole Language Theory" was infamously misguided to assert that students will learn to love reading when we provide them with the freedom to choose their own texts, it is a fallacy to set our sights on getting students to love writing for its own sake. This goal is fruitless. Writing is important because it makes learning visible, it enables deep thinking as it is cross-disciplinary, it empowers us to create, and it can facilitate the development and preservation of our mental health and identity. This list could go on, but can be put more succinctly when simply drawing our attention toward the LLM's for which I am critical. Without the vast array of our original thoughts, information, and research freely available online, these LLMs could not be as great as they truly are. Meta's alleged stealing terabytes of copyrighted works to produce Llama is a compelling example.

This is not to come across as some hermitic Luddite. Students absolutely hate writing and will avoid it at any cost, and they should harness any resource in their grasp to shepherd their authentic thoughts into reality. This does not mean they should be enabled to skip this process altogether though.

A means or an end?

We've been lucky that our most vital institutions so far have rejected the prospect of using generative AI in lieu of their own critical thought, and we need a system to get students to understand the same.

To this end, I present a straightforward litmus test for AI use: Am I using AI as a means or an end to a product?

This line of thinking is relatively intuitive. Whether it is brainstorming, outlining, researching, or writing, teachers act as the lighthouse which students use to steer their work toward the right direction, and this simple test can help maintain the goal posts for teachers and students alike.

Reviewing the sources from the ChatGPT search function? That's a means.

Putting in your rough draft to have Mistral completely re-write it to sound more academic? That's an end.

Through effective modeling and formative work that leave the door open to using these tools, our little test can slowly become a routine, nagging in the back of students minds while they work. When consistently applied to classroom activities, it becomes far more clear what use AI can present, and where it does not belong, in the same way we teach students the uses of Wikipedia for academic research.

But this test does not occur in a vacuum, and building this routine is clearly not a cure to the determined-to-plagiarize pupil. Educators ought to continue to utilize current best practices for AI in their teaching, while keeping in mind this overarching philosophy of where our AI line is. Whether it is using Google Assignments to see document version history, placing constraints on assignments to keep students honest, creating scaffolds and check-ins that task students to explain their evidence and reasoning throughout a project, or requiring a degree of specificity that would be difficult for an LLM to adhere to at a high fidelity, this line of thought should guide you in making your own pedagogical decisions.

Students like our dismayed learner from the beginning of our tale are by no means uncommon. They're young, they're human, and they just want to get our assignments done. Though this mistake is dire, it is not the end of the road. We'll have them complete their narrative with a tutor who can work directly with them. As the semester marches on, we can reinforce what we want to see in their work, and why we want to see it. With clear reasoning, expectations, and routines, we can implicitly guide all learners to become independent thinkers who use technology as a tool, not a crutch. Though Moore's Law carries us forward, clear habits of thought like this will develop an intuition which will prepare us to respond to our rapidly changing world as it passes before us.