Spoiler alert: there's nothing in there.
Reactions to OpenAI’s newest iteration of a language-stringing algorithm are many and varied. It’s the end of high school English. No, it’s not. Let’s ban it from schools. Let’s have teachers use it. Okay, maybe not. No, it will be just like using a calculator. Maybe we should take it to court, because the whole things is based on massive plagiarism and unauthorized use of creator’s work. It might be good for getting a job, but not for getting a date. And we can expect a whole new range of reactions when OpenAI inevitably rolls out a for-pay version of ChatGPT.
There’s even a technological counterattack, as companies roll out technology designed to detect software-composed writing.
But the bottom line is that ChatGPT and other composition algorithms like it are not going away. It’s not the apocalypse, but it’s not a nothingburger, either. More than anything else, it’s a chance for teachers who use writing in the classroom to do some careful, thoughtful reflection.
The teaching of writing has too often involved teaching students to follow an algorithm. Your essay will have five paragraphs; start the first one in with a sentence about your main idea, then fill in three paragraphs with supporting ideas, then wrap it up with a conclusion. Fill in each paragraph with three-to-five sentences stating your idea and then providing supporting details. Or, answer the essay prompt on the test by rewriting the prompt as an opening thesis statement.
Call it a format or a template or an algorithm. Schools have taught students to build structures, to craft sentences to order, to treat writing as a dance of performative hoop jumping. Schools have taught students to assemble essays to satisfy algorithms for judging their writing—algorithms that may be used by either humans or software, with little real difference.
Now we can clearly see that this type of writing can be completed, not brilliantly, but passably well, by a piece of software that literally has no understanding of what it’s writing. If this kind of writing can be done by a machine that doesn’t have a single thought in its head, what does that tell us about what we’ve been asking of students.
Some teachers have told students that the basic building block of an essay is a sentence. But software can make a sentence, pretty much to order. I argued throughout my career that the basic building block of an essay is not a sentence, but an idea.
Teaching with composition algorithms is comforting for both students and teachers. Students are anxious to know exactly which hoops they need to jump through to get that grade, and teachers who are less confident about teaching fuzzy art and craft of writing find in algorithmic writing instruction a set of concrete, easily measurable factors to use for grading. But the unfortunate side effect is that teachers end up grading students not on the quality of their end product, but on how well they followed the teacher-required algorithm.
Composition algorithms, sentence mechanics, and grammar all have their place. They are the tools of the writer’s work, and just as a builder needs to know the difference between a hammer and a nail gun and a screwdriver and a tape measure, young writers benefit from knowing their tools. However, we judge builders by the quality of their work, not by how well they hold a hammer.
ChatGPT won’t end any kind of writing instruction that shouldn’t have been ended. Writing for real, and not for performance, is personal and grows out of the writer’s thoughts and ideas, and that’s where writing instruction should be focused. Grow assignments organically out of class discussion and interest. Grade essays on how well the student communicated an idea, and not on how well they filled up the five paragraph template.
ChatGPT and other language software can help. They can be used to test the quality of a prompt; if the software can create a passable essay, then why ask students to do it?
And ironically, as more and more people play with ChatGPT, it’s becoming evident that to get a better result from the program, the user has to put the kind of detail and thought into their instructions that should be used for writing the essay themselves. ChatGPT is a dynamic demonstration that if you approach an essay by thinking “I’ll just write something about Huckelberry Finn,” you get mediocre junk. Better thinking about what you want the essay to be about, what you want it to say, and how you want to say it gets you a better result, even if you’re having an app do the grunt work of stringing words together.
Most writing problems are really thinking problems, and thinking problems are the problems that software is completely incapable of solving.
If a teacher’s concern is that students will “cheat” with ChatGPT, the answer is to give assignments that are personal and focused on thinking. We don’t have to teach students to follow a writing algorithm any more; there’s an app for that.
