Why Human Writing Is Worth Defending In the Age of ChatGPT
Naomi S. Baron on the Detrimental Impact of Artificial Intelligence on Literacy and Cognition
A specter is haunting the landscape—the specter of generative AI. First came fears that student cheating would explode, plus that artists and actors would be unemployed. Then the ante was upped: Some of the very technology’s creators warned that AI’s potential risk to humanity as we know it was on par with pandemics and nuclear war.
This cascade of angst was triggered by the launch of ChatGPT by OpenAI in November 2022. Among its wiles, the bot is best known for its prowess at churning out prose. On an existential level, does it matter if AI writes for us? For a non-kneejerk answer, we need to make a serious stab at understanding how writing affects us as people.
Most fundamentally, it changes our minds and brains. The classicist Eric Havelock argued in Preface to Plato that development of writing and concomitant spread of literacy in Archaic Greece, even in limited circles, enabled the flowering of Greek philosophical thought. Writing facilitated reflection, logical thinking, and production of tangible texts to foster rethinking.
While Havelock’s argument for historical mind change has had its critics, it’s incontrovertible that literacy changes our brains. Thanks to modern neuroscience, we know the brain is “plastic,” meaning it is capable of reorganizing its structure or laying down new pathways, depending upon our physical or mental activities. London cabbies with “the Knowledge” of thousands of routes, streets, and landmarks have larger posterior hippocampi (the area responsible for physical navigation) than control groups. And people who are literate have different brains than those who aren’t. Using MRI scans, Stanislas Dehaene taught us that adults who only learned to read and write later in life increased the density of white and gray matter in brain areas involved in reading.If writing helps us think, what happens when we surrender the process to AI?
The literate brain empowers us to use writing as a canvas for witnessing our thoughts. Recall Flannery O’Connor’s much-quoted remark that “I write because I don’t know what I think until I read what I say.” Her sentiment is hardly unique in literary annals. You’ll find it echoed by Horace Walpole, E.M. Forster, Arthur Koestler, George Bernard Shaw, William Faulkner, and, of course, Joan Didion (“I don’t know what I think until I write it down”).
If writing helps us think, what happens when we surrender the process to AI? We risk becoming cognitively and expressively disempowered.
Start with AI as editor of text we write ourselves. Spellcheck and basic grammar and style edits from Microsoft Word are old news. But newer tools like Grammarly and Microsoft Editor (both now infused with OpenAI’s GPT models) are at once more potent and perilous, especially for less confident writers.
Can AI edits be trusted? Some bones of contention are small beer. Word continually instructs me where to put my commas and to be more concise. Sentence-initial “Finally” must be followed by a comma; replace “in the near future” with “soon.” A matter of personal choice, you say. Yet sometimes the advice is flat-out wrong. When I wrote
“However we might define ‘good’ writing, it’s more than acing checklists.”
Word scolded me for not inserting a comma after “however.” Sorry, Word. A comma doesn’t belong there, since the adverb “however” is modifying “define,” not the whole sentence.
More troubling still was the “inclusiveness” flag that Microsoft Editor blandished at me when recently I wrote that Rameshbabu Praggnanandhaa (nickname: Pragg) was “the new Indian wunderkind,” describing a sixteen-year-old who astounded the chess world in early 2022 by beating five-time world champion Magnus Carlsen. Microsoft’s style cop warned that “This language may imply bias against indigenous populations” and suggested I substitute “Indigenous” or “Native American.” Yes, there are indigenous populations in Pragg’s home country. But they’re indigenous to India, and I doubt that’s how Pragg would characterize himself.
I knew to ignore Word’s counsel. From Pragg’s Hindu name alone, not to mention the long tradition of playing chess in the country that invented the game, it was obvious to me what “Indian” meant. But writers for whom it wasn’t self-evident might have ended up substituting “Native American,” which would have been absurd.
Beyond the issue of trust is that of personal writing voice. Take predictive texting (an early version of generative AI). Harvard research has shown that when we use predictive texting, our vocabulary tends to become more concise and less interesting. The philosopher Evan Selinger warns that this AI shortcut encourages us “not to think too deeply about our words” and to “give others more algorithm and less of ourselves.” Describing predictive texting, a student in one of my studies complained that “I feel like the message I sent is not mine.”
AI as editor and author is with us for the long haul. Writers need to make peace with the language genie, while holding fast to the opportunities for thinking and creating that writing bestows on us. Since we all have different aspirations when it comes to writing, the peace we broker must be individual. Approaching our own negotiating tables, here are two considerations to keep in mind.
First, beware of deskilling. Like retaining a foreign language, writing takes continuing practice. Nir Eisikovits, a philosopher, warns that the biggest near-term threat of AI is that it will lull us into degrading “abilities and experiences that people consider essential to being human.” I used to ask my students, What do you know when the internet is down? Today’s concern: Are you still capable of writing if an AI editor or text generator is unavailable?
Second, recognize your level of commitment, especially when something written bears your name. It’s child’s play for the likes of GPT-4 to concoct emails, blog posts, and article summaries. Is that OK with you? If you opt to view use of AI as a collaborative venture, how much are you willing to accept its algorithmic edits or text it has created out of predictive whole cloth? Research by Shakked Noy and Whitney Zhang found that ChatGPT reduced the time humans needed for a writing task, plus improved the quality of the end result. What’s more, 68% of study participants were content to submit ChatGPT’s initial output without doing editing of their own.
That’s handing over the keys.
The bulk of writing that most of us produce skews impersonal. It’s the everyday tasks of emails and memos, maybe churning out news stories or school assignments, even writing a guide to successful ChatGPT prompts. AI has already proven itself highly adept at such endeavors.
But human motivations for writing run deeper. We write to look outward, as with literary works that convey our perspective on the human condition. We write to look inward, including to find out what we’re thinking. We write for personal release, be it a diary entry or angry letter to an employer. All this writing is grounded in human sentience, of which AI has none. AI has no drive to better people’s lives, impart what it is thinking, or convey emotion.
When it comes to weighing commitment, also remember that both as a form of personal expression and as an art form, writing is a craft. How we chose our words and sentences is as vital as the meaning they convey. I keep harkening back to a passage in Francine Prose’s Reading Like a Writer:
“Writing…[is] done one word at a time, one punctuation mark at a time. It require[s] what a friend called ‘putting every word on trial for its life’: changing an adjective, cutting a phrase, removing a comma, and putting the comma back.”
If we cede to AI final say about words and even commas, we jeopardize more than artistic pride. We risk convincing ourselves that in the name of efficiency, it’s harmless for AI to assume ever wider swaths of what we previously would have written ourselves.
Don’t get me wrong. I’m no troglodyte when it comes to partnering with AI in the writing enterprise. Rather, my counsel is not to lose sight of the precious tool that writing offers us for shaping our minds and brains, for articulating our own ideas, and for sharing them with fellow humans.
Naomi S. Baron’s Who Wrote This?: How AI and the Lure of Efficiency Threaten Human Writing is available now from Stanford University Press.