To AI or Not to AI - There is No Question

IN MY SYLLABUS, I restrict the use of AI in completing assignments at every stage in the writing process, and I provide my students with two explanations for this policy.

The first explanation is a metaphor: Using AI in an English class is like going to the gym and paying somebody else to lift weights for you. It's a waste of tuition money; the purpose of college is to learn and grow, and that requires the friction that comes with challenging assignments, just as strength and endurance come from the friction of exercise and cardio.

The second explanation is a little more abstract: Most students come to college with the idea that writing is about "rules that you follow," but the truth is that writing is about actions and their consequences. Every word, every sentence, is a choice that students are free to make, or not, and as such, learning how to write in college is about learning to use the abundance of writing choices at our disposal. This training data is essentially the purchased or stolen writing of actual people scraped from the internet and meshed together to form an algorithmic prediction of what generative AI programs are designed to regurgitate, which are customer satisfaction tokens. Programs like ChatGPT are designed to plagiarize thousands of writers and stir together billions of words and spit out an answer designed to please potential customers, regardless of factual accuracy.

This is engrained in our culture, but creative decision-making is also the core of our evolution as a species. We evolved larger brains that not only store memories, but connect them, and writing is the intentional process of making connections. When we sit down and make a series of decisions about how to write to a professor requesting an extension, we use creative thinking as a form of problem-solving as our neurons leap into action. In essayist Kazim Ali's words, "Chemically, it is invented in the brain. Thought is matter."

Generative AI is not currently capable of replicating this process. Tech writer Ed Zitron describes generative AI as "probabilistic, meaning that [AI programs] don't know anything, they're just generating an answer (or an image, or a translation or a summary) based on training data."

This training data is essentially the purchased or stolen writing of actual people scraped from the internet and meshed together to form an algorithmic prediction of what generative AI programs are designed to regurgitate, which are customer satisfaction tokens. Programs like ChatGPT are designed to plagiarize thousands of writers and stir together billions of words and spit out an answer designed to please potential customers, regardless of factual accuracy.

Outsourcing any part of the writing process to AI means voluntarily refusing to think about a subject. If I asked ChatGPT to give me an outline for an essay about Stephen King's first novel, Carrie, the program would run a gargantuan search through the data that has been stuffed into it, which in this case would include book reviews, scholarship, film reviews, biographies of Stephen King, student essays and probably some miscellaneous information about real people named Carrie. It would then condense those ideas, produced by real people, into a completely unoriginal and entirely plagiarized list of the most common themes and interpretations of the novel, which would then direct me toward a thesis based on already-existing arguments.

I want my students to learn to think for themselves because college is an environment where they will be challenged and will encounter many different viewpoints. This brings me back to the abstract explanation. As a teacher, grammar and syntax are far less important than students learning how to access, express and navigate their own original ideas. There is more educational value in writing a clunky, error-filled rough draft that expresses an original opinion than to have a polished and organized final draft regurgitating what everybody else already thinks.