Composition Studies 51.1 (2023): 155–161
Meta-Writing: AI and Writing
Aimée Morrison
C
an I just say? And is one of my favorite coordinating conjunctions. It’s so
deceptively simple, sitting there in between two things, just connecting
them—but the sneaky thing about and is that you never quite know what
kind of connection it is making until you gure out what kind of connec-
tion it is making. And can be used to connect two parallel items (oranges and
lemons), to indicate sequence (go and sit), to indicate addition (and baby
makes three). It can also, weirdly, be used to double up on identical words:
to indicate progression along a state (I am getting hungrier and hungrier), to
intensify (I have been writing for ages and ages), or—God help me, what a
joker this little word is—to ag how one word can be itself in more than one
way: “Aimée is writing and writing this piece.”
1
e prompt “AI and Writing” is an evocative formulation: it juxtaposes
two concepts without indicating the nature of the relationship between them.
What is the relationship between them? is prompt opens up ideas. I begin
to imagine possible relationships, branches of inquiry or argument, and now
I’m curious. “AI and Writing” is a provocative formulation, too: the relationship
of AI to writing is currently the matter of heated debate, often polarized. e
agnostic attitude connoted by the and feels dangerous and urgent: this is no
time for neutrality! is is existential stu! I experience a strong, possibly nega-
tive reaction, as if the phrase is goading me with its impassivity: my heart races
and my palms start to sweat. So now I’m also a bit agitated, in the part of my
brain where facts meet values, and ideas meet structures. When I encountered
the prompt “AI and Writing,” then, I got excited: my aect, certainly, is “very
enthusiastic and eager,” and there is something elemental happening, too, more
akin to the way excited is dened in physics as “an energy state higher than
the normal or ground state” (Apple Dictionary app), like a crackling cloud in
a rumbling sky when the hair on your arms starts to stand up just before the
discharge of lightning, like the Beastie Boys rapping “I’m rolling down the
hill, snowballing, getting bigger / An explosion in the chamber, the hammer
from the trigger” (Beastie Boys).
is is what writing feels like for me, every time: surprising and joyous
and confounding and allusive and elusive and annoying and impossible.
Writing: “AI and Writing”
e passage above took me two weeks to write, but also 20 minutes to write,
and then a week to edit.
156 Composition Studies
What I did was this: I chewed on the prompt of “AI and Writing” in my
head for about a week, started free writing on my computer and scribbling on
bits of paper at random times, became distressed by how upset the topic made
me and the sheer number of reasons that it upset me, discussed some ideas
with friends, collected and read a blizzard of op-eds and journalistic pieces,
(re)wrote many hundreds of words into the stubs I had started, and became
overwhelmed. I then capped this stage o with the advanced self-soothing/
avoidance techniques of expelling my best one-liners on Twitter, practicing
particularly complicated passages of Debussy for many hours on the piano
accompanied by a relentlessly dinging digital metronome, summoning a
martini, and taking to my bed early with the hard sudoku on the New York
Times web app.
What I did was this: I woke up, still sulking, ate Shreddies doused in
almond beverage, in bed, and suddenly realized I had just spent two weeks
enacting in my process what I had been trying to grab at with my prose. I
opened my laptop, then Pages, then the dictionary app. I looked up coordi-
nating conjunction, and then I looked up and, and then I looked up excited,
and then I looked up evocative, and then I looked up provocative, and then 20
minutes later 20 percent of this piece was written. en I scribbled to myself:
“Beastie Boys,” “Duckworth dog bone,” “stupid sci- baegab magic Clarke,”
and dove head rst into some brand new internet holes where I spent the next
hour getting giddy on ideas.
What I did was this: Repeatedly snuck back into those paragraphs for 20
seconds or three minutes at a time, fussing the words, or the pace, then snuck
back out again, for the next week.
What I did was this: wRiTInG and ‘writing’ and writing and writing.
Generated Response: “AI and Writing”
Articial intelligence (AI) has become increasingly prevalent in the eld of
writing, with numerous applications and tools available to assist writers in
various aspects of the writing process. Some of the ways in which AI is used
in writing include:
1. Grammar and spell-checking: Many word processing programs and
online writing tools incorporate AI-powered grammar and spell-
checking features that can help writers catch errors and improve the
quality of their writing.
2. Content creation: AI-powered content creation tools can help writ-
ers generate ideas, develop outlines, and even create entire pieces of
content based on specic keywords or topics.
Meta-Writing 157
3. Language translation: AI-powered language translation tools can
help writers translate content from one language to another, mak-
ing it easier to reach a wider audience.
4. Text analysis: AI-powered text analysis tools can help writers ana-
lyze the eectiveness of their writing by providing insights into fac-
tors such as readability, tone, and sentiment.
5. Writing assistants: AI-powered writing assistants can provide writ-
ers with suggestions for improving their writing, such as alternative
word choices or sentence structures, helping to improve the overall
quality of the writing. While AI has the potential to be a valuable
tool for writers, it is important to note that it is not a substitute
for human creativity and critical thinking. Writers should still be
mindful of the context and purpose of their writing, and should
use AI tools as a supplement to their own knowledge and exper-
tise, rather than relying on them entirely. With the right balance of
human skill and AI-powered assistance, writers can produce high-
quality content that eectively communicates their message to their
intended audience.
“Indistinguishable from Magic”
In 1973, the science ction writer Arthur C. Clarke oered up his ird
Law: “Any suciently advanced technology is indistinguishable from magic,”
which is the kind of thing science ction writers like to pronounce, and op-ed
writers like to quote (Clarke; Dunn; Karpf). Certainly, it is not dicult to
nd popular writing on generative AI that leans heavily into awe, wonder, fear,
and other forms of “magical” thinking. I think this is misplaced: a generative
AI is mostly lightning-fast computational grinding. e “writing” it spits out
is essentially the sum of brute force data scraping and storage, brute force pro-
cessing power, and some clever procedures that have as their ultimate purpose
a passable mimicry of written human language. Essentially, ChatGPT can
largely be understood as the code version of the “innite monkeys theorem,”
in which it is surmised that, given enough time, an innite number of mon-
keys bashing away at an innite number of typewriters will, without muse or
intent or an MFA, reproduce, say, the works of Shakespeare. It’s not magic;
it’s just grunt work made workable by the sheer scale of resources thrown at it,
with the resources in the case of generative AI being of the computing rather
than simian variety. A Large Language Model (that’s what ChatGPT is) is
programmed to calculate a plausible-sounding (not “correct,” please note)
text output in response to a prompt. It’s just faster monkeys if those monkeys
had been told what task they were meant to accomplish and been given a
158 Composition Studies
massive subset of human writing to cut, paste, and recombine from instead
of randomly smacking at individual typewriters keys.
What’s really magic, actually, is writing. Writing is speech by other means,
and both more and less than speech. Writing is an act of communication, of
persuasion, of instruction, of exploration. What an amazing, unlikely, hopeful
thing writing is: to think beyond the present moment, to develop a system of
inscription and abstraction to catch and save and share ideas, to imagine an
audience, to desire to reach them. What a human thing to do! I wish we spent
more time being amazed by writing. I wish we spent even some tiny fraction
of the time, curiosity, eort, and money that we are willing to spend to teach
machines to imitate writing on teaching human beings how to write. Mostly,
we are enmeshed in systems where we nd ourselves removed from the pos-
sibility and opportunities for amazement, instead nding ourselves harried,
under-resourced, and worried that maybe writing is over and our jobs along
with it. is moment of reckoning, though, might also be the evocative and
provocative prompt that lets us see What Writing Is And What’s It’s For with
more clarity and power—and to ght for it.
Writers versus “Writing”
Consider my response to the prompt “AI and Writing,” versus what Chat-
GPT generated. I hope you found my writing engaging and surprising. I was
trying to capture the giddiness of discovery, the play of ideas and words, the
means by which the act of writing or even thinking about writing is the most
powerful form of active learning I have ever experienced, to show rather than
tell. I’m reading bits out loud to my teenager and laughing my ass o, and
then re-writing things to put the beats where I want them, changing words,
moving sentences. I am trying to open a little space in your brain to see
the question of AI and Writing the way I see it. As a 750-word essay on AI
and Writing for a rst year academic writing course? I would probably earn
a pretty lousy grade, if we’re being honest. ere is the personal pronoun.
Emoji. Expressive typography. Non-standard paragraph structure. No clear
thesis. Contractions. Colloquialisms. e Beastie Boys are not an academic
source. It is a surprising response to the provided prompt, and “surprising” in
academic writing courses is usually not celebrated. In fact, my writing seems
almost willfully insubordinate, self-sabotaging, as if I’m “just looking for a st
/ to put [my] face in” (Beastie Boys).
e AI-generated response, by contrast, is a marvel of correct, mild-
mannered, balanced, objective prose. It has a clear introduction and a sum-
mary conclusion. It makes clear points organized sensibly into a numbered
list, which is consistently formatted and pleasingly parallel in construction
and theme. No careless errors or shifts in tone or wild swings at stray pitches.
Meta-Writing 159
No split innitives. e impression it gives is of smooth condence and com-
petence, the soothing quality of authority that oozes from abstract nouns in
sentences crafted in faultless Standard English. If you read it with an eye to
content, however, you might notice: the response is a bit repetitive, and it’s
also internally contradictory in ways it doesn’t seem aware of.
• Are “writing” and “content” the same thing?
• What is the dierence between “grammar and spell-checking” and
“text analysis”?
• Is the goal to improve writing or produce more writing faster?
• What is a “writing assistant”?
ere seems to be some sort of unaddressed power struggle between human
writer and AI content creation hidden inside abstract nouns: e items are
grammatically but not semantically parallel. Nevertheless, it’s the kind of ex-
pository writing—real talk—that would get a high grade in most high school
classrooms and many university ones. Hypnotized by form and tone, it’s easy
to miss that this says almost nothing concrete, or valuable, and doesn’t really
cohere. Or perhaps we don’t really look: one of my own colleagues, hearing
me rant about the insipidity of AI writing, told me, “If my students produced
insipid writing that was grammatically correct? I would be delighted. A+.”
“Delight Me”
Every year I hand out syllabi to rst year university students afraid to write lit-
erally anything at all, who enter my classroom as if walking towards their own
executions, who experience reading and writing as a rigged game with secret
rules applied arbitrarily by a series of stern authority gures drawing blood
and draining the life force from their GPA with runic spells cast by red pens:
frag, agr, colloq, red. ey want to see the rubric. ey want to know the
formula. ey want grammar worksheets for extra credit. ey want to know
exactly what they have to put on paper and how to control the outcome. ey
want to know what ideas they are meant to be arguing. ey do not want to
free write; they look physically pained when brainstorming. Writing is an
obstacle between them and their future, a future without writing in it. ey
are resentful. ey will tell me on the rst day of class: I don’t want to be here;
I’m bad at writing; I’m never going to need this. Just tell me the right answer,
and I’ll do it. e conclusion I’ve come to is this: my students think that
good writing is producing an error-free response to a prompt—that there’s
an ideal answer that the prompter already knows, and the goal of writing is
to approach that secret answer as near as possible, with the least amount of
risking being wrong. ey think this because it has been, actually, much of
their experience of writing in school contexts, which is what they consider to
160 Composition Studies
be writing. In many cases, we have been training them to try to write like a
machine would. Smooth, bland, correct, formulaic, compliant, voiceless, not-
wrong. ChatGPT does it better, but it’s never been worth doing: this is why
students hate writing it and you hate grading it.
In a rst year writing-intensive English course on Digital Lives, I had a
student come to my oce after the rst assignment, to ask where he had “lost
grades.” Computer science and engineering students ask me this question very
frequently. He had done quite well, and I told him so in my standard response.
“But how,” he asked, “if I wanted to get a higher mark? What would I have
to do?” I surprised myself by answering: “Delight me.” And then he surprised
me: he kept coming to my oce, kept talking through with me what could
distinguish writing that got the job done from writing that was delightful. How
the way we express the idea—through pacing or vocabulary or register—makes
up a good part of persuasion and eectiveness. I told him to read more, to
notice the writing, to mark where he was surprised, to feel where the beats
landed. He would need to risk a big idea and try to nd the way to understand
it by writing his way in, taking his reader with him. He kept pushing, and I
kept pushing. By the end of the term, he was getting grades in the mid 90s.
It was not so much that he was submitting better writing projects, although
he was, but that he was really developing a sense of what writing can be and
do. We were both delighted.
Writing without Writers
Generative AI pretends that there can be writing without writers, which is
as nonsensical as suggesting there can be swimming without swimmers, or
breathing without breathers. e advantage that human writers have—be-
yond our crystallized knowledge, contextual understanding, desire to con-
nect, and gift for linguistic and syntactical innovation—is that we seek, pro-
duce, and discern meaning. e mirage of suciently advanced AI blinds us
to the fact that it is nothing more than an imitation game, an innite number
of monkeys, a stochastic parrot. e generated text is boring, or derivative,
or both, and often factually incorrect. When I write, it’s because I have some-
thing to say, and I know that if I do it persuasively, I can change some part of
the world in some small way. It’s because I love endlessly looking up words I
already know, to disrupt my reliance on received understanding. It’s because
I know I will emerge with so many more ideas than I started with, and I nd
that exciting. It’s because the way I’m going to write is going to sound like me
and that’s how that diers from what others would write is what makes up
the scholarly conversation. My students deserve to feel like this too, but they
don’t. Most of my writing instruction is now oriented toward teaching my
students that their own ideas are worth expressing, that they’re not “writing
Meta-Writing 161
up results” so much as “writing into understanding,” that their language is
valuable, and their experiences relevant, that I am a real reader ready to learn
from what they write. I mostly try to nurture in them the passion, purpose,
and perseverance that Angela Duckworth names as the core components of
“grit”—that quality that motivates and sustains doing hard and ambitious
things. A writer who does not care about a topic, and sees no point in writing,
is never going to write anything worth reading. at isn’t just a poor imita-
tion of some poorly understood ideal instance—since ChatGPT is so good
at producing that kind of text, we can with condence release our students
from that burden.
And where might we go then?
Notes
1. “And” is a conjunction, I feel, that, like, gets diérence in a fundamental way.
Works Cited
Beastie Boys. “Looking Down the Barrel of a Gun.” Paul’s Boutique. Capitol, 1989.
Clarke, Arthur C. “Clarke’s ird Law on UFO’s.” Science, vol. 159, no. 3812,
Jan. 1968, pp. 255–55. DOI.org (Crossref), doi.org/10.1126/science.159.381
2.255.c.
Dunn, om. “Physicists Identify New Quantum State Called ‘Magic.’” Boing Bo-
ing, 1 May 2023. boingboing.net/2023/05/01/physicists-identify-new-quantum-
state-called-magic.html.
Duckworth, Angela. Grit: e Power of Passion and Perseverance. First Canadian edi-
tion, Collins, 2016.
Karpf, David. “Money Will Kill ChatGPT’s Magic.” e Atlantic, 21 Dec.
2022, www.theatlantic.com/technology/archive/2022/12/chatgpt-ai-chatbots-open
ai-cost-regulations/672539/.
