Carly Schnitzler is a PhD student and teaching fellow in English and Comparative Literature at the University of North Carolina at Chapel Hill studying the intersections of contemporary poetry, experimental composition, and digital infrastructures.

* * *

When @poem_exe tweets:

“a hazy day

lovers part

in a bead of sweat”

as it did last fall, the account’s over 18,000 followers do not imagine a writer sitting behind her computer, condensing image and linguistic creativity into a pseudo-haiku and publishing it for the world to see (@poem_exe, 3 Nov). We know @poem_exe has not experienced a hazy day, has not parted from a lover, certainly has not sweat a bead of sweat. The entity directly behind @poem_exe is a bot, run from a rented server, auto-generating these captivating short poems five times a day, for the past five years. Of course, the bot has a creator, real person Liam Cooke, an Irish software engineer living in Australia, who used the translated haikus of another real person, 18th century Japanese poet Kobayashi Issa, as the source of his poetic database (Cooke). Issa’s poems form the corpus that Cooke’s algorithm uses to enact a digital version of the OuLiPo technique created by French mathematician and poet Raymond Queneau in the 1960s, exploring the infinite possibilities of poetic verse that can be produced under certain structural constraints (Cooke).

@poem_exe and other artistic Twitter bots like it are sites of poiesis. Poetic creation can happen without a human author directly engaged in the act of creation. This is not an alarmist or sinister claim—robots may take our jobs, but they will not take our poetry. Inserting these bots into the processes of poiesis complicates and evolves the role of human authors and readers in a way that is rhetorically valuable for contemporary poetry. This kind of algorithmic poetics dislocates and manipulates authorship in a way that is conceptually significant for bots and for digital poiesis. Examining the evolved roles of authorship in these bots through the rhetoric of the bots themselves blends new materialism and rhetoric in the digital, initiating a discussion of rhetorical ontology previously disconnected from the digital realm.

The @poem_exe bot digitally adapts Raymond Queneau’s compositional technique in his foundational OuLiPo work A Hundred Thousand Billion Poems in which each line from a set of ten sonnets is printed on its own separate strip, encouraging recombination to form any one of literally a hundred thousand billion possible poems. Spearheaded by a group of mostly French mathematicians and writers whose ranks included Marcel Duchamp and Italo Calvino, the OuLiPo movement enacted “systematic, self-restricting means of making texts” that now offers both structural and ideological compatibility with bots enacting poiesis (Gallix). Instead of being limited by Queneau’s cardstock and set of ten sonnets, though, @poem_exe becomes a site of poiesis ad infinitum, under a similar set of structural constraints.

In the @poem_exe bot, there are a few authorial gaps worth exploring, breaking down the mechanics of and responsibilities for creative production. First, the poet-engineer, in this case Liam Cooke, identifies a database of content to manipulate—the translated corpus of haiku written by 18th century Japanese poet Kobayashi Issa. He then prepares the database for his intervention, engineering an algorithm to scrub extraneous information and organize the data in a way that makes sense for his future bot to use, separating out and identifying the words and syntax of each line of haiku. He sets this aside and creates the algorithm of poiesis that will do the work of making for the bot. For @poem_exe, that means he codes together natural language processing packages to determine syntax for constructing each line of haiku, computational functions for determining statistically common or “seasonally relevant” terms or phrases (Issa had a particular affinity toward snails, and this is reflected in the work of the bot), and a randomizing software to compose lines of rough haiku, in a relatively standardized three-line form (Cooke). Cooke also creates a copy-editing algorithm to control for syntactical regularity and consistent orthography, which is run on top of the algorithm of poiesis. He then rents servers and makes a Twitter account. The database, scrubbing algorithm, algorithm of poiesis, and copy-editing algorithm are layered, run through the servers, pushed through to Twitter’s API, and then individual haiku are published on the Twitter account, @poem_exe. Enigmatic and infinite possible haiku are realized, composed, and published five times a day.

For @poem_exe and the other artistic Twitter bots, the poet-engineer acts as a divine gesturer, determining the terms and processes of poetic creation, but leaving the actual work of production and creative outputs up to the bot. In clearing the space for poiesis and determining the scope of it, the poet-engineer allows for production without producing final creative outputs. It is also important to note that the poet-engineers can—but have no obligation to—override their bots’ authority over final outputs, acting as a curator for creative outputs that are not aligned with algorithmic intention or simply do not make sense.  With the component parts of the bots disassembled, then, we can begin to examine their rhetorical structure. I will start first in a traditional rhetorical mode, by examining the speaker-audience-purpose triad and its structural insufficiency, then move to a rhetorical new materialist perspective to better capture the properties of authorship unique to the bots.

Traditionally, as rhetoricians Diane Davis and Michelle Ballif maintain, the triad around which a rhetorical situation revolves is composed of  “a knowing subject who understands himself (traditionally, it is a he), his audience, and what he means to communicate” (Davis and Ballif, 347). This speaker-audience-purpose triad can be applied to the poetic work of the bots, but in doing so, the humanist application eclipses the authorial nuances and gaps between poet-engineer and the bot as a generative site of poiesis, disregards the readerly interaction with the creative outputs, and misreads the purpose of the exchange.

Under the speaker-audience-purpose model, the bot has no creative force and cannot, because it is not a knowing, self-comprehending human subject, occupy the space of ‘speaker,’ despite it being the producer of final creative outputs. So, the poet-engineer has to shoulder the full weight of the speaker role, disregarding the authorial gaps present in the process of creating the content of the tweet. As poet-engineer, in this rhetorical situation, Liam Cooke is the sole ‘speaker’ of the poems tweeted by @poem_exe, choosing an extremely circuitous route to tweet ‘his’ poems. Followers of the Twitter account are the natural audience in this triad formulation, but both the unique premises of the bot accounts and the mediation of social media platforms like Twitter complicate the relationship of Twitter followers as audience to Cooke as ‘speaker’ on an account like @poem_exe. The ‘purpose’ juncture of this rhetorical situation, tentatively applied to artistic bot accounts, is where the triad collapses, and the influence of the bot cannot be overlooked. Cooke sets out, as a speaker of sorts, to experiment with OuLiPian techniques of poetic composition via the mechanism of the bot, not to tweet out recombined pseudo-haikus himself. Followers of the @poem_exe account follow the account because a bot, and exactly not a knowing, self-comprehending human subject, is the one producing creative outputs in the form of tweets.

Barbara Warnick and David Heineman, in their updated version of Rhetoric Online: The Politics of New Media, use Maurice Charland’s constitutive rhetoric and Kenneth Burke’s theory of identification to address the shortcomings of traditional rhetorical formulations for digital environments, particularly around social media. Though their work is primarily focused on the intersection of political and digital rhetoric, these more nuanced rhetorical theories get us closer to the true operation of the authorial mechanisms in Twitter’s art bots, particularly in light of their digital context. Charland’s prescient 1987 article “Constitutive Rhetoric” problematized the “ontological status of speaker, speech, audience, topic, and occasion” as not being neatly delineable by their agency, but decoupled from human agents, who can occupy or partially inhabit any number of these roles as a single individual in a given rhetorical situation (Charland, 148). His work has had sustained relevance in discussions of digital rhetoric, since his formulation of the speech act is dynamic in its impact on the multiplicity of agents that surround it—“a user’s identity…is determined in specific ways by the text” (Warnick and Heineman, 104).

Further complicating the purpose of a speech act, Burke’s theory of identification “allows for the possibility of self-persuasion or creative participation with a text” (Warnick and Heineman, 97). This shift in and messy-ing of purpose then has an effect on author (rhetor) and audience: “For Burke, rhetorical identification is something that works when an audience feels that they share the same worldview and life experiences as the rhetor. The rhetor, therefore must use the tools available in order to facilitate this feeling of mutuality” (Warnick and Heineman, 98). Parsing this further, this conceptualization distinguishes speech act or “text” as separate from both rhetor and audience (Warnick and Heineman, 97). Also implied in this is the distinction of the initiation or gesturing of a speech act from the gesturer and the difference between the initiation of a speech act from its completed state as text. The “tools” used to facilitate the “mutuality” between rhetor and audience in Burke’s theory of identification, then, take on a much more active role (Warnick and Heineman, 98). We see here the onus of rhetorical purpose beginning to shift (at least partially) off of the shoulders of the human author, and onto the tools at their disposal.

This opens up space for a bot to exist in rhetorical situations, as a vibrant rhetorical tool, with bequeathed creative capacity set into motion by poet-engineer as divine gesturer. Here, I would like to make my own distinction in line with this rhetorical formulation, between gesturer and rhetor. In the territory of the artistic Twitter bot, the human poet-engineer is the gesturer that produces the rhetorical situation (and its subsequent terms and rules) in which the nonhuman bot can then act as rhetor producing creative output in tweets, enacting poiesis. This rhetorical situation empowers both human gesturer and bot rhetor, while also mapping well onto how new materialism conceptualizes of its capacity to oppose the binaries of “both transcendental and humanist traditions” (Dolphijn and van der Tuin). The term “gesturer” is used intentionally, as a rhetorical adaption of the new materialist “gesture,” in which “the immanent gesture of new materialism is transversal rather than dualist” (Dolphijn and van der Tuin). The emphasis on transversalism (as opposed to dualism) is a key tenant of new materialist thought, led by theorists Rick Dolphijn and Iris van der Tuin in their seminal monograph on the field, New Materialism: Interviews & Cartographies. Transversalism relies on the Deleuzian conceptualizing of subjects, objects, and events as relatively equal monads in a pervasive series of relations to one another. This theory opens up an infinity of possible relations, between disciplines, schools of thought, humans, texts, and everyday things, imposing loose structural constraint to produce a hundred thousand billion (or more) rhizomatic relations, not unlike the practice of Queneau’s OuLiPo works that offer inspiration for some of the bots in question.

Rhetoricians Scot Barnett and Casey Boyle in their collection Rhetoric, Through Everyday Things blend rhetorical theory and discussions of materialist-oriented ontologies in productive response to the “unsettled relations between things,” offering emerging precedent for combining rhetoric and new materialism (Barnett and Boyle, 9). As their foundation, Barnett and Boyle “take ontology to be the pervasive relationality of all things—the means by which things come into relation and have effects on other things in ways that resonate strongly with existing and emerging understandings of rhetoric” (Barnett and Boyle, 9). In other words, this is a rhetorical take on transversality. Their work primarily contends with the rhetorical authority of  physical, and not digital, everyday things (like typewriters and dinner plates and dragonflies), though they do acknowledge that “digital and media technologies offer fertile ground for discussion of rhetorical ontology” (Barnett and Boyle, 13). This work explores that “fertile ground,” oriented specifically towards the poetic territory Twitter bots can occupy (Barnett and Boyle, 13).

In setting out to answer questions of authorship in artistic Twitter bots, we need to rely on the transversality between and the monadic ontological status of the collaborative authorial efforts of human gesturer, bot rhetor, and an active human follower to enact poiesis. The humanist binary that underpins the speaker-audience-purpose triad fails to capture the complexity of authorial relations happening in the bots, prompting this move to rhetorical new materialism. Conceptualizing of poetic authorship as a series of material relations “dismantles” this humanist binary and rhetorically enables poiesis as it happens in the bots (Dolphijn and van der Tuin).

Taking these new materialist ideas in tandem with compatible rhetorical theories, like Charland’s constitutive rhetoric which adopts its own prescient version of new materialist ontological relations, allows for a fuller picture of authorial relations in the bots. Using Charland’s idea that text can inform user identity as a springboard, we can work from an exemplary tweet outward to its various authors, as a way to demonstrate the joyful complexity of authorship in the poetic territory of the bots. @mothgenerator is a bot that knits together poet-engineer, bot, and follower in an incredibly compelling interaction that merges linguistic and visual elements, in addition to entangling authorial relations. The algorithm of poiesis corresponds visual components that compose illustrations of moths with language and syntax sourced from tweets directed at the bot. The bot turns the linguistic content of the tweets directed at it into a unique illustration of an imaginary moth—I tweeted at @mothgenerator, instructing it to make a ‘William Carlos Williams’ moth and this is what it created, using the linguistic content of ‘William Carlos Williams’ (@mothgenerator, 14 Nov).

The tweet here, as with the traditional rhetorical model, remains at the center of the rhetorical situation. But, instead of speaker, audience, and purpose neatly revolving around it, human gesturer, bot rhetor, and active follower form their identities and relations around the illustration of the ‘William Carlos Williams’ moth and each contribute authorship in different entangled capacities (Charland). The human gesturer, or in this case Loren Schmidt and Everest Pipkin, the two poet-engineers behind @mothgenerator, ‘author’ the terms and processes of poiesis, they stake the claims and legislate the territory of it. The bot rhetor ‘authors’ the individual creative outputs, enacting poiesis on this claimed territory. The active follower here ‘authors’ input, catalyzing the poetic reaction. The active follower input for @mothgenerator is explicit and linguistic, making it a good example of the authorial engagement followers have in the bots. Even for bots that do not algorithmically rely on follower creativity, though, they still rely on an active followership. The social structure of Twitter is predicated on these sorts of active, opt-in engagements, making it a natural media for this poetic territory.

The different authorships coalescing in the bots transverse “humanist and transcendental binaries” and the subject-object divide and, in doing so, reimagine authorship as an expansive, transversal series of relations (Dolphijn and van der Tuin). This new materialist ethos allows for the rhetorical separation of identity from function, in a blending of Charland and Burke’s rhetorical theories. Authorship does not belong to a single agent, human or non, but rather is used differently and strategically by each interacting entity upon the interface of poiesis that is the Twitter bot. Without any one agent fully responsible for final creative outputs, the territory of the bot is expansive and generative, providing structure for infinite creation, unique to its digital form. The bot as territory is a site of poetic empowerment, for all of its authors. It calls upon human gesturer, bot rhetor, and active follower to become entangled in the refreshing, functional work of poiesis. Through this entanglement, we see the gaps between authorship in the bots that are equally as present, gaps in which linguistic serendipity lives.

Works Cited

@mothgenerator. Twitter, https://twitter.com/mothgenerator.

@mothgenerator, “the the william carlos williams moth moth.” Twitter, 14 Nov. 2018, 5:54 a.m., https://twitter.com/mothgenerator/status/1062705323717394432.

@poem_exe. Twitter, https://twitter.com/poem_exe.

@poem_exe, “a hazy day/lovers part/in a bead of sweat.” Twitter, 3 Nov. 2018, 10:21 p.m.,   https://twitter.com/poem_exe/status/1058952246283558912.

“A Brief Guide to OULIPO.” Poets.org, Academy of American Poets, 3 Oct. 2015, www.poets.org/poetsorg/text/brief-guide-OuLiPo.

Barnett, Scot, and Casey Boyle. Rhetoric, Through Everyday Things. University of Alabama Press, 2017.

Burke, Kenneth. A Grammar of Motives, and A Rhetoric of Motives. World Pub. Co., 1962.

Charland, Maurice. “Constitutive Rhetoric: The Case of Thepeuple Québécois.” Quarterly Journal of Speech, vol. 73, no. 2, 1987, pp. 133–150., doi:10.1080/00335638709383799.

Cooke, Liam. “Poem.exe.” Poem.exe, 2014, poemexe.com/.

Davis, Diane, and Michelle Ballif. “Pushing the Limits of the Anthropos.” Philosophy & Rhetoric, vol. 47, no. 4, 2014, p. 346., doi:10.5325/philrhet.47.4.0346.

“Electronic Literature Volume 3.” Electronic Literature, Electronic Literature, collection.eliterature.org/3/index.html.

Gallix, Andrew. “OuLiPo: Freeing Literature by Tightening Its Rules.” The Guardian, Guardian News and Media, 12 July 2013, www.theguardian.com/books/booksblog/2013/jul/12/OuLiPo-freeing-literature-tightening-rules.

“OuLiPo.” Poetry Foundation, Poetry Foundation, www.poetryfoundation.org/learn/glossary-terms/OuLiPo.

Parrish, Allison. “Allison Parrish: Recent and Selected Work.” Allison Parrish: Recent and Selected Work, 2018, portfolio.decontextualize.com/.

Queneau, Raymond, and Lionnais François Le. Cent Mille Milliards De poèmes. Gallimard, 1961.

“The Transversality of New Materialism.” New Materialism: Interviews & Cartographies, by Rick Dolphijn and Iris van der. Tuin, Open Humanities Press, 2012.

Warnick, Barbara, and David Heineman. Rhetoric Online: the Politics of New Media. Peter Lang, 2012.

* * * * * * * * * * * * * * * * * * * * *

* * * * * * * * * * * * * * * * * * * * *