WHAT DOES AI KNOW ABOUT GOOD WRITING?

By Kristen di Gennaro

Many of us have used AI for various reasons. Google searches use AI to find sources in response to a question. Currently, Google provides an overview instead of sources. This fully formed response uses generative AI, technology widely available since 2022. Generative AI has led to widespread panic, especially among college teachers who see changes in students’ writing. Gone are the sentence fragments, run-ons, subject-verb discrepancies and other signs of a writer more familiar with spoken than written language.

AI has made grammatical accuracy irrelevant for evaluating writing. For some, this is a welcome wake-up call, forcing teachers to look beyond the low-hanging fruit of grammar as a stand-in for judging writing. Discriminating against writers unfamiliar with prescriptive grammar rules has become harder.

Fears about students using AI reminded me of the shift, years ago, from handwritten placement exams to typed responses on campus computers. A colleague of mine insisted on disabling the spellcheck feature on each computer, believing that students’ ability to spell correctly, under timed conditions, contributed to accurate (and meaningful) writing assessments. By this logic, how many of us currently misrepresent ourselves when we allow our computers and phones to correct our spelling?

Another colleague and I recently used generative AI to draft conference slides from a chapter we had written. At first glance, the slides looked great: clear phrases accompanied by colorful images. Looking closely, however, revealed talking points lacking a logical structure, impressive sentences hiding faulty coherence and bulleted lists failing to show meaning beyond each phrase. Details and main ideas meshed together without distinction. And the images? They bore no relationship to the content. In short, AI’s version offered pretty slides with disconnected ideas. If writing is reduced to grammar, AI wins top grades. But writing is more than just a collection of sentences.

AI generates language, not original ideas or critical thinking. It produces better output in response to better prompts. Good users read the output and refine their questions. Writers using AI well engage in the process of writing by asking clear questions, critically evaluating responses, asking more questions and revising responses to meet expectations. Writing teachers have been preaching for decades about process over product; with AI, we can see that process unfold. Instead of accepting superficially impressive responses (like our initial conference slides), students can work through AI’s responses and develop writing that is not just grammatically

accurate but also factual, well organized, logically argued and connected to course content and goals. With less focus on grammar and more on meaning, students might even think more deeply while teachers read more fluently.

I’m not a fan of AI, but I’m a realist. We can resist AI and argue against its widespread use, but we can also celebrate how AI has given us tools to shift our focus from product to process in writing instruction. And unless we think spelling and grammar are the best ways to judge people, AI might even succeed in reducing linguistic discrimination.

Disclosure: Generative AI was not used in writing this column. All mistakes are my own.

Please share your thoughts on these and other language matters with your local linguist at languagematters@mattersmagazine.com.

---

Kristen di Gennaro is an associate professor of English at Pace University, where she teaches courses in linguistics.