Writing as Thinking: What We Lose with Generative AI
Writing as Thinking: What We Lose with Generative AI
A meditation on what happens when we let generative AI write for us
I’ve been pondering what Chris Lewis, the Catholic illustrator at BARITUS, said about AI removing the sacrificial dimension in making art. If you let generative AI make the drawing for you, it may be easier to get a finished product. But by the same token, you miss out on the sacrificial dimension of making the art altogether. You also never learn how to draw.
As per Lewis, the process (not the product) defines the artist.
If you don’t believe him on this point, just watch him make this splendid illustration of St. Anthony’s temptation.
Of course, the artist must take an interest in how the finished product will turn out. Art is not a good occupation for those unconcerned with the final result. For many, actually finishing an essay, book, sculpture, song, drawing, etc. can seem like absolute torture. And I think this is Chris’ point. The difficulty in bringing out a form from some sort of recalcitrant material can be experienced as a sacrifice.
Not everyone wants to experience sacrifice in every given situation, though, and that is perfectly understandable. For example, if you use writing to sell things, then you may simply want ChatGPT to write the copy for you. The goal is not to get lost in the process but rather to make a widget (in this case, written copy). Indeed, you don’t really care about the process. I regularly use generative AI to create images for these Substack posts. If I wanted to get better at drawing, I would need to stop doing that.
However, if you use writing to think, as many do, then generative AI will hinder your capacity to think in writing. This much should be obvious. It isn’t that you won’t think at all. You just won’t have the form of thinking that writing affords.
We think in language, and we think differently in the written word than in the spoken word.
In this post, I want to reflect on how, with generative AI, we trade process for product. In essence, we trade one form of thinking for another. I want to home in specifically on how generative AI affects writing since I tend to do more of that than anything else. And many of the readers of this Substack also have an interest in writing.
Three Facets of Writing as Thinking
Here, I’ll rely on Fr. Walter J. Ong’s essay “Writing is a Technology that Restructures Thought” to flesh out how writing changes our thought processes. In particular, the writing process grants three things to those who actually engage in it: distancing, depth, and discovery.
Distancing
We think differently with distance from a situation. The more we write, the further distance we can put between ourselves and a situation. Per Ong:
"Writing is diaeretic. It divides and distances, and it divides and distances all sorts of things in all sorts of ways” (Ong 1986, p. 36).
Diaeresis simply means division, a taking apart.
What is one advantage to writing? It distances thought from a context. If you’re angry or emotional, you may need distance from a situation in order to think reasonably. Time affords distance in this manner, as the Stoics knew. But so does writing.
Elsewhere, Ong writes:
"Alienation from a natural milieu can be good for us and indeed is in many ways essential for fuller human life. To live and to understand fully, we need not only proximity but also distance” (Ong 1986, p. 32).
When we use generative AI, we lose this experience of distancing, this slowing down of thought and methodical backing away from a situation in order to process it.
Thinking in writing allows us to consider situations and relationships otherwise, often in a manner that allows us to process them more analytically and with less emotion.
Depth
We think deeper in writing. Indeed, forcing ourselves to translate what another says by putting it into our own writing confers understanding, a point made by Ahrens in his book How to Take Smart Notes.
The longer the piece of writing, the more opportunity for nuance in our thought process. Two people writing long-form essays contradicting their points will have greater opportunity to analyze and expand upon ideas than people tweeting at one another.
Even Plato benefited from writing. It allowed his thought to go deeper. Per Ong, writing transformed and influenced Plato's thought process. Writing enabled Plato to engage in "protracted, intensive linear analysis” (Ong 1986, p. 29).
When we write, we can see our thought, which (if you really think about it) is a very strange thing to see.
"Between knower and known writing interposes a visible and tangible object, the text” (Ong 1986, p. 38).
When you fix a thought in front of you, you can return to it with greater ease. Constantly repeating a thought to yourself requires effort. It is fatiguing. You’ve probably experienced this fatigue in an argument where you’re waiting to get your word in.
In a sense, writing allows you to hang your thoughts in mid-air. You can allow them to dry. You can fly them up the flagpole. But, in any case, your thoughts benefit from the fresh air, from the ink meeting the paper.
Of course, journaling is a way of seeing your thought processes. How you feel about yourself, your relationships, and your immediate milieu might otherwise pass invisibly out of existence forever. Journaling allows you to fix yourself into an object that you can analyze. Such fixing can be negative, of course (see my post on reification). But, depending on the context, fixing yourself into writing can also afford you with deeper ways of thinking about and processing a situation.
With digital media, thought processes thin out, a point made by Nicholas Carr with all his work on “the shallows.”
Carr coined the term “the shallows” to talk about what the Internet was doing to our brains: it was leading to shorter attention spans and a decreased capacity for entertaining long and/or complex arguments.
You could make the case that gen AI is doing the same thing but for writing instead of reading.
In short, generative AI does for writing what search engines and the Internet broadly speaking did for reading.
Discovery
Perhaps most of all, using generative AI to write removes the need to develop your own writing process, which in essence is your own thought process. If you never develop your own thought process, you never really develop your own ideas.
You never discover anything that you can call your own.
When we write, we complete research, outline, etc. We learn as we draft. But if you no longer have to draft because gen AI will simply spit out a complete piece of writing for you, you no longer learn along the way. You no longer find those serendipitous and surprising things along the way that drafting enables you to find.
Certainly, I think it possible to incorporate gen AI into an existing writing workflow (e.g., getting a preliminary idea about a topic, producing research to follow up on and validate, formatting citations, etc.). But it is also possible for it to take over the entire composition process itself.
Drafting requires you to constantly drag your many supporting points back to your one overarching point. This constant dragging is both sacrificial (see above) and liberating. You can perceive the purifying of your thought as you do this over and over, as you make your piece about one thing.
“What do you think?” versus “How do you think?”
So, generative AI transforms our thought process because it transforms our writing process.
Now, just a word in closing about why all this matters.
If you want to object to generative AI, you will struggle to do so if you focus your attention on the product. Generative AI makes nice products. The images are creepy and surreal every now and then, admittedly, sure. But many a person has been fooled already by a “deepfake” or a piece of generated text.
We cannot object to generative AI because it is artificial. Writing itself is artificial. Playing a musical instrument is artificial. Save for vocals, all music comes from some sort of technology or machine (Ong 1986, p. 33).
Thus, the capacity to hedge an effective critique against generative AI will stem directly from your ability to demonstrate what we lose, especially in terms of how we think.
Every technology enhances and obsolesces something. By outlining how we think in writing, we can effectively begin to acknowledge what we lose when we give up writing for ourselves.
Perhaps most ironically of all, we will think about generative AI differently (and perhaps in a shallower way) if we use generative AI to think about generative AI. For that complex and most important topic, we need something else. Can you guess what it is?
Thanks for reading. Please like, subscribe, and comment. If you’re feeling really energized, make 100 copies of this post and hand it out to random strangers. Make sure they know it is coming from TeachDelightMove.com.
Giving credit where credit is due: If I recall correctly, my mentor in grad school used to talk about how we need distance for relation. I think Martin Buber makes the same point.