AI is coming for poets’ jobs — or so researchers at OpenAI, most famously known as the maker of ChatGPT, would like you to think. In a paper published in 2023 examining the “Labor Market Impact Potential of Large Language Models,” 4 researchers (three of whom were employed by the AI company) posited that “Poets, Lyricists and Creative Writers” are among the occupations most highly exposed to being automated by AI models like that underpinning ChatGPT.

While it may be hard to imagine poetry being a prime target for automation, this is already in motion. The Poem Booth, designed by Dutch design studio VOUW, emerges from the notion that yes, writing a poem is work, and yes, in human hands, that work is performed much too slowly. 

The tech object — an actual photo booth that snaps a picture of users, then programmatically writes a few rhyming couples about the portrait in a matter of seconds — uses a fine-tuned version of ChatGPT to wow users with rapidly penned, personalized poems. Thus far, The Poem Booth has been touted as a crowd-pleaser at conferences and conventions, though VOUW also puts it up for rent (in tech parlance, we might say “Poem-as-a-Service”) for the price of EUR 950 for 6 hours.

But are The Poem Booth’s couplets any good? Do they have the capacity to move, to spark action, to speak from one heart to another? That is to say, are they even poetry?

My research paper will investigate the extent to which The Poem Booth and other commercial attempts to automate the poet (see Google’s “Bard” chatbot or Anthropic’s “Haiku” and “Sonnet” AI models) run antithetical to the spirit of poetry as a craft. I will also explore the potentialities of the web as “a space of poiesis” — Loss Pequeño Glazier’s term — in which new forms of poetry can take root. 

Glazier’s words appeared in the early 2000s, when the web looked vastly different, both in its ethos and in its capabilities. I will approach this same space from the vantage point of the mid-2020s, in a post-ChatGPT age. There is now a digital poetics at the margins of the internet that seeks to wake readers up — not only to their own existence, but to their habituated experience of a web that has become too heavily trafficked and commoditized. I will use recently published examples from online journals like the html review, Taper, and Crawlspace in my formulation of a potent, born-digital poetics that rescues language from the maw of large language models.

Play serves as the undertow of this new digital poetics: browsers that play text, code that puts text into play, readers-as-interactors (to use Nick Montfort’s term) that play with and within the text. In an extension of Ian Bogost’s procedural rhetorics, this digital poetics seeks expression through a flash of double-rhetoric — that of computational procedure and written text. In particular, digital poetics’ embrace of kinetic and multimedia structures privilege a simultaneous, integrated act of persuasion.

A couple weeks ago, I purchased Stardew Valley, an “open-ended country-life RPG” that first came out in 2016. The farming simulation game’s premise is that you, the player, have become disillusioned with the trappings of modern life — the game starts with a cut scene of your character working in a gray cubicle, feeling increasingly forlorn — and have decided to move to Stardew Valley, where you’ve inherited an overgrown but otherwise appealing plot of land from your deceased grandfather.

I have only spent a combined 45 minutes or so playing the game, so I can hardly speak to the full extent of how the game works or the avenues of exploration it offers. But already, it’s clear to me it is making some sort of statement, and encouraging me to consider this statement through the gameplay.

Given its potential as an object of meaning-making and discourse, I thought it’d be perfect to run it through some of Ian Bogost’s ideas on procedural rhetoric from his book Persuasive Games. 

Persuasion through process

One of the most interesting mechanics I’ve encountered in Stardew Valley is its borders and boundaries. My character has a plot of land that he owns, but he lives near others in the Valley who also own plots of land. These are typically bordered by fences, but they are more gestural than useful, as there will be gaps in the fences where you can easily enter their property. You can then go right up to their door and walk inside. Once inside, you can explore the items in their house, inspecting their kitchen, bookshelves, etc. But if you try to enter their bedroom, the game will prevent you from doing so — until you’ve befriended the homeowner. For now, you don’t know them well enough yet to enter their bedroom. 

Like the fences with their porous gaps, houses occupy a kind of hybrid private/public space in Stardew Valley. The implication is that the world is full of such boundaries. And my character (and therefore, I) comes into contact with this innate procedurality, like an IF ELSE statement in programming — IF my character isn’t friends with this person, don’t let him into the bedroom; ELSE let him in. 

Simply by nature of interacting with this condition, I’m brought deeper into the situation — it’s clear to me I should go befriend this person so I can unlock this door; OR if I don’t really care about what’s inside, maybe I won’t waste time trying to get to know them. The mechanic opens up a choice. This is a hallmark of modern discursive rhetoric, which features, Bogost writes, “the effective arrangement of a work so as to create a desirable possibility space for interpretation” (19). My character becomes an agent in exploring, altering, and interpreting that possibility space. 

Along these lines, I would have been interested to see more process built into the portion of Stardew Valley’s opening scene, where my character is a cog in the machine of corporate operations. This is built in as a cut scene — a video that I can either watch or skip. But I’m not actually made to interact with this situation, and this is a missed opportunity to help me better understand why my character is so disillusioned. This is where procedural representations can actually do more heavy lifting (in terms of persuasion) than a simple video.

Vividness and Stardew Valley’s graphical choices

Charles Hill’s “comprehensive continuum of vividness,” reproduced by Bogost, places “actual experience” as the most vivid form, followed by “moving images with sound,” “static photograph,” and more, all the way to “abstract, impersonal analysis” and finally “statistics” (34-35). Bogost revises this list to include procedural representations, specifically those with “with high process intensity and with meaningful symbolic representations” (35), in the second spot, i.e., the most vivid form after actual experience. 

Bogost later posits video games as one of the more robust forms of procedural representations, typically featuring “more process intensity than other computational media” (44), greater expressiveness, and a high degree of interactivity (45). Bogost frames the best video games as being very vivid, not because they’re so immersive but because they effectively create meaning by abstracting away from the real world. 

Stardew Valley is a prime example of this concept at play. Its graphics are markedly unrealistic — their colorful, 8-bit aesthetic makes it feel more like an early Pokémon game than any of the ultra-realistic (and ultra-commercial) games of the 21st century (GTA, Call of Duty, etc.). 

But this supports the game’s rhetorical underpinning of ditching modern society. It does away with hyper-optimized, uncannily striking visuals to place you more deeply in a life away from all that. It inverts the idea that “visual fidelity implies authority” (Bogost 49), as if to say that top-down authority is what ruined modern life in the first place (at least in the sense that the industrial drive toward production / commoditization alienates the modern worker from their labor).

Discourse in and around the game

In Bogost’s definition of procedural rhetoric, a key aspect of the rhetorical angle is the ability to spur discourse — not only to advance the game’s position or argument, but also invite subsequent responses, objections, modifications, etc. from players. Bogost’s example of “The Grocery Game,” for instance, features “a flourishing community of conversation” (39) built on message boards, where players discuss strategies to win, as well as share stories about their experiences. Per Bogost, “…the availability of this forum facilitates active reconfiguration of the game’s rules and goals” (39-40), which supports the vibrancy of the discourse surrounding the game and its players.

In Stardew Valley’s case, look no further than Reddit to find many fun and insightful conversations surrounding how people experience the game. It expands beyond gameplay and into fan art devoted to the game’s characters, reinterpretations of the game’s music, and even accounts of real-life interactions with strangers that were sparked by the game.

All of this makes Stardew Valley into a thoroughly expressive, discursive artifact of digital culture — one that, in my eyes, persuades its players through its built-in processes and mechanics, while also giving them the space to push back, both via in-game decisions and off-platform conversation.

I’ve run regularly for about 15 years now, but never in a competitive or team setting (like cross-country or track). I prefer to run solo. And I prefer to be unencumbered by devices: no arm band, running watch, or belt equipped with small bottles of water and energy goos. 

The most I’m really willing to do is sign into the Strava app and press “start” when I start my run, and “stop” when I stop it. For the uninitiated, Strava is a platform for tracking exercise and connecting with others through exercise in a social network-style community (instead of “liking” someone’s post, you give “kudos” to their run/walk/cycle route/etc.). I first tried it when training for a half-marathon, with the hope that it would help me stay accountable to my loose training plan. 

After using it for a few years now, I can’t say I’ve ever been encouraged by the app to really stick to anything. It doesn’t send me push notifications telling me to get outside or play high-tempo music to push me to run faster. It doesn’t read out my mile times as I’m running, like Nike+ did the one time I tried it. In my opinion, all of this works to Strava’s benefit. I like that it doesn’t try to “nudge” me too much toward certain behaviors. 

However, what I appreciate most is that it affords space for me to share context about my run after the fact, and I get to share this with friends. These elements (the contextual, the social) are ancillary to the run itself, and to the app’s supposed reason for existing, but they allow me to experience more joy in and around the run. This quality hints at what Miguel Sicart describes in “Playing The Good Life” as the “expressive” nature of play, which allows players to “play freely with and within [the] boundaries” (235). 

The app doesn’t necessarily invite this kind of play, but it doesn’t foreclose it, either — the way Nike+ (another running app, and therefore a helpful point of comparison) does. As Sicart writes, “Users of Nike+ can only make sense of their experience of running through the data manipulation tools that the system affords, such as the limited description field for the run where users can describe the quality of the run. Nike+ in this way limits its users’ agency over the practice of running” (233). 

Strava, on the other hand, allows you to write whatever you want about the run. It doesn’t limit the description field to a limited range of possible answers (e.g., via a drop-down menu) or even to a short text description. It prompts you with “How’d it go?” and then you can write as short or long of a response as you’d like (or leave no response at all). 

Here is where I like to get playful in my responses. I’ve described a 2-mile run as “Slobbery” (I was running with my dog, and I attached a photo of him with his tongue out). I’ve chalked my fast pace up to the fact that I was trying to finish in time to take advantage of “happy hour” at a bar. These addendums help attach the activity of running to the context in which it occurs — thereby undermining the kind of detachment that most gamification architectures achieve.

This helps me feel less like a “user” and more like a “player.” A game’s players, according to Jennifer Whitson, “bring their modes of play along with them. This play includes joyful explorations and tangential detours/détournement. It also includes counterplay, both of which complicate the surveillance projects that constitute corporate gamification endeavors” (“Foucault’s Fitbit,” 355). 

While being welcoming to a certain kind of playfulness, Strava is, at the end of the day, a business — and a booming one. It made nearly $300M in revenue last year, primarily off subscriptions for premium features. But its users aren’t the only audience it’s going after. Strava has a separate business line called Strava Metro, through which urban planning organizations and city governments can use Strava’s rich data on athletes’ movements to inform city development and safety. While this data is currently offered free of charge, it reflects the obvious value inherent in its proprietary data on the over 100M athletes that use its app. 

As a private company, Strava isn’t held to the same reporting standards or scrutiny as a public company — its founder has even stated that the company is in no rush to go public given “a lot of disclosure” requirements. It’s not hard to imagine the company diverting user data toward more commercial ends if and when it needs to access new revenue streams. This happens all the time in corporate settings — as Whitson notes in the case of RescueTime, which defended the release of an employee monitoring service for employers by saying: “Revenue and profit are king and we can’t expect to focus on free/consumer audiences forever” (352).  

Look no further than Strava’s current job openings: a job description for a Senior Director of Data, for instance, says that anyone hired into this role should plan to “Uncover measurable insights about our global community of athletes, both for internal and external consumption; turn volumes of data into product features, partnership opportunities, and other strategic interests.” The phrase “other strategic interests” leaves the door open to all kinds of possibilities. 

But for now, users seem content to play within the borders of the Strava app — including seeing where and when they can bend the rules to their favor.

In deciding how to spin Lewis Carroll’s Alice’s Adventures in Wonderland into a game, my group took cues from Carroll himself in dispensing with rules or decorum in service of more playful ends. That is to say, we broke the text open, stepped inside, and turned it upside down. 

From the start, we were drawn to putting the text into some kind of motion. We briefly considered altering how a reader reads the text (e.g., manipulating the materiality of the text by playing with its shape, size, etc.). But we wanted to push this further by rethinking the reader’s role altogether. We decided to turn the reader into a player with the capacity to influence the directions and outcomes of the text. And so we built a game out of it.

In our initial brainstorming, we kept circling back to the Cheshire Cat as a wholly unique character, even in a book full of unique characters. It is one of the few characters to repeat across numerous scenes, to the point that the cat becomes a kind of (maddening) spirit guide to Alice. Because its intentions, desires, and goals are never made clear in the text, we felt the Cheshire Cat would be a perfect foil for the player, who could similarly engage with Wonderland and its inhabitants without a defined set of aims.

It was important to maintain the Cheshire Cat’s moral ambiguity in the incentives and goals of the game play. Accordingly, the game never insists that you should prefer one outcome (e.g., keeping Alice in Wonderland) over another (e.g., waking her up). The player gets to interpret and enforce one or the other modes of intervening in the story.

To distribute the Cheshire Cat’s presence throughout the narrative, we decided to rewrite portions of the original text to interpolate the cat’s presence in scenes where it was previously absent. I primarily focused my efforts in this area, writing the Cheshire Cat into multiple “decision points” that would serve as forks in the branched narrative of the game. I also converted the narrative into a more game-friendly format, which often meant sacrificing Carroll’s circular brilliance for something more direct and fast-paced. (We didn’t want to lose the player’s attention — we’d all had our own experiences playing video games with interminable cutscenes.)

It would be fair to ask, If you’ve changed the plot points, and now also the writing, is this even Alice’s Adventures in Wonderland anymore? While I can’t easily answer that, I can say for certain that the original text has been dislocated — from its time, its place, and its objecthood. By transforming it from a linear story into a “narrative world” (to use Janet Murray’s term), we’ve changed the terms of the story and the assumed expectations of roles like “reader” and “author.”

I also wanted to briefly discuss the group element of this project. We all came at this with very different skill sets and areas of interest. And it was only in collectively moving toward the slippery areas between our respective competencies and capacities that we were able to create this wonky, mischievous piece of digital ephemera. Now that the game is live, any player that enters the game from this point forward will also participate in co-creating the game space. In a way, they’re going down the rabbit hole with us.

This summer, as a result of moving, I reinvested in tending to my vinyl record collection. It had lapsed for a decade because I lacked the space necessary to accommodate the bulk of a record-playing setup — the player itself, the receiver, the speakers, a collection of records that threatens to overgrow whatever container holds it. But now I’d moved into a space that could better support the hobby.

Backing up a bit: I was born in the CD era, a few years after the medium had eclipsed tape cassettes as the way to own music. CDs could hold more songs and they were less fussy. But by the time I was really coming into my musical consciousness, the internet was taking over, and legally dubious digital spaces like Napster, Limewire, and The Pirate Bay offered ways to access a much wider range of music for a fraction of the cost. And from there: streaming. Pay a nominal fee each month, forget about questions of legality, and listen to (almost) anything.

So why vinyl? I believe it has to do with play, especially in the more expansive sense that authors like Bogost and Huizinga use in describing it.

The ritual of collecting and listening to records is highly ceremonial. It requires going to a record store (or perhaps a thrift store, stoop sale, or estate sale) and thumbing through overstuffed bins of faded album covers. This is one of the practice’s playgrounds. The other playground is at home, in front of the record player, where the ceremony culminates in sliding the record from its plastic or cardboard protector, taking pains not to get fingerprints on the grooves, placing it on the platter, and then delicately lifting and dropping the needle along the record’s edge.

These are signs of respect to a self-imposed order. As Bogost writes in Play Anything, “Fun comes from the attention and care you bring to something that imposes arbitrary, often boring, even cruel limitations on what you — or anyone — can do with them. Worldly limitations impose a new and welcome humility, for they force us to treat things as they are rather than as we wish them to be” (13-14). The worldly limitations of listening to vinyl — the necessity of getting up from dinner to flip the record, for instance — stands in stark contrast to streaming, which eliminates nearly all friction.

Listening to vinyl is less about satisfying a momentary urge (“I want to hear this song this very instant!”). In fact, as Huizinga writes, play “interrupts the appetitive process” (Homo Ludens, 9). When I listen to vinyl, I am subordinating my own desires to the limits of the present moment. I feel closer to the things themselves, the materials of my game — approaching a greater sense of the “worldliness” that Bogost praises as a hallmark of play (7).

I think this inverse relationship — between fun and constraint — is partially responsible for the strength of the vinyl industry right now. Although diminished compared to streaming, record-buying has made a huge comeback, and this rebound has roughly tracked with the growth of streaming over the last decade. This tells me I’m not the only listener who, when given what they want without limits, opts for some anyways.

This feeds into one of the core debates of record-collecting: whether buying records that were recorded digitally (as opposed to using analog methods) counts as real record-collecting. I admit this is a myopic view and an attempt to keep people out of a hobby that should be enjoyed widely. But it speaks to the need for rules in maintaining the illusion of a game. Even after being pressed into vinyl, any trace of the digital starts to break the spell.