@FlyOnTheWallBot {Twitter bot}

The Twitter Bot as Oulipo Machine? 

Procedure (n): The fact or manner of proceeding with any action, or in any circumstance or situation; the performance of particular actions, esp. considered in regard to method; practice, conduct.

I have built a Twitter bot: @FlyOnTheWallBot whose name is “Casual Observer.” It describes itself as this: “I am the fly on your wall. I see you and I feel your feels.” Modeled after Mark Sample’s @SaveHumanities bot, which begins each tweet with the syntax “To save the humanities, we need” and then culls from recent tweets that include the starting phrase “we need” to finish the sentence and outputs it in a new tweet, the Casual Observer searches Twitter for the most recent tweet with the phrases “I see you” and then searches for the most recent tweet with “I feel like” and recombines the results in a new sentence which the bot then tweets. Here are a few examples:

Although I’ve been on Twitter for several years, I feel as though I am just getting a sense of how different communities and networks of people make use of it. For me, Twitter provides a space where I can promote my current thinking and projects, links to cool things I’ve encountered on the Internet, and interact with like-minded poets, friends, and graduate students as well as witness the interactions of scholars and writers whose work I admire. When I’m poking around the search function, however, I can easily eavesdrop on people I do not know at all for shared hashtags or any given key term. This last part, which feels very much like eavesdropping, is part of what intrigues me about Twitter as a social and cultural phenomenon and was the initial impetus for this bot. I wondered how I might simultaneously investigate and replicate for others that odd and slightly awkward feeling of witnessing public confessions of encounter and emotional resonance when those are present (not that the people I follow tend to use Twitter as a confessional space). Imagining the bot as a fly on your wall puts into place a narrative frame through which the resultant tweets can be read. The idea that this fly is a “casual observer” immediately calls one to question whether or not the kind of digital eavesdropping the bot has been programmed to perform can ever be casual. Indeed, I’d argue, it is always critical. In short, because this bot’s algorithms cull from live search results of Twitter users, the resulting tweets provide poignant, albeit often humorous, commentary on our public behaviors, at least in the Twitter-sphere.

Procedure & Poetics, or the Affordances of a Poetics of Inquiry

The experimental poetry group the Oulipo (Ouvroir de littérature potentielle) has emphasized the playful and generative nature of procedural and constrained modes of composing by seeing language and mathematics as analogous systems of meaning making. Employing algorithmic procedures such as n+7, where every noun in a chosen text is replaced with the next noun 7 words down in the dictionary, is the key example of an Oulipo procedure, but it also includes other kinds of constraining devices, such as the lipogram (famous in La Disparition Perec’s 300 page novel composed entirely without the letter “e”). This kind of pre-determined constraint and the recombinatory procedures that determine how an existing text might be (re)arranged are at the heart of the Oulipo. It is also, I argue, at the heart of this kind of bot.

Warren Motte in the introduction to Oulipo: A Primer of Potential Literaturelays out the following:

Simplifying grossly, one might postulate three levels [of constraint]: first, a minimal level, constraints of the language in which the text is written; second, an intermediate level, including constraints of genre and certain literary norms; third, a maximal level, that of consciously preelaborated and voluntary imposed systems of artifice. No text can skirt the minimal and remain readable; perhaps no text can wholly avoid the intermediate level. But it is the maximal level that concerns the Oulipo: this is what they refer to in using the word “constraint”; this is what characterizes their own poetic production and, consequently, the model they propose to others (11, emphasis in original).

The importance I’m placing on this kind of inheritance is the Oulipo’s attention to the ways in which all language is constrained; in Oulipoan practices and in Twitter bot construction the constraint is foregrounded, but as is all forms of constraint, whether self-imposed or derived from generic or other conventions. What is a bot but a “consciously preelaborated and voluntary imposed system of artifice”? As such, these works also depend on reader interaction to complete the system of meaning-making. Given the social and cultural constructions of norms that inform our reading of any text, (re)combinatory methods resist and revise those conventions by making fissures visible in the potentially nonsensical text output. It works both with and against our expectations for language delivery.

The recombinatory Twitter bot does this as well. The syntax is based on a reasonable sentence we might encounter in many different kinds of texts, in particular the tweet. “I see you… and I feel like…” is merely observation and reflection. Outputs like “I see you get your own happiness and I feel like giving up” is a reasonable (albeit sad) sentiment that could have been expressed by one person. But outputs like  “I see you lookin at ma curves coke bottle and I feel like I’m the only one who doesn’t like to be called babe” or “I see you on my timeline undefined and I feel like I’m affiliated with the illuminati” make sense but are also just slightly odd enough that a singular person most likely wouldn’t have said them. Even whatever glitch is spitting out the word “undefined” instead of the thing my tweet code can’t interpret makes for an interesting phrase: “I see you on my timeline undefined”.

Procedures such that the Oulipo employ are also dependent on reader interaction and interrogation.  Indeed, “the concept of potential writing corresponds to that of potential reading. Faced with this conclusion, some may feel the Oulipian generosity is overshadowed by Oulipian brashness: the radical valorization of the status of the reader, rather than a gift, may seem a date” (Motte, 21). Some may think of the Bot and feel unhinged by its potential slippages in convention, or frustrated by the proximity to an understanding that can never quite be reached. But as Motte continues, “serious and plaful intent are not mutually exclusive in the Oulipo’s work: they are, on the contrary, insistently and reciprocally implicative” (21).

In François Le Lionnais in Lipo: First Manifesto (1962) he emphasizes the mission of the Oulipo:

In the research which the Oulipo proposes to undertake, one may distinguish two principal tendencies, oriented respectively toward Analysis and Synthesis. The analytic tendency investigates works from the past in order to find possibilities that often exceed those their authors had anticipated… The synthetic tendency is more ambitious: it constitutes the essential vocation of the Oulipo. It’s a question of developing new possibilities unknown to our precedecors” (Le Lionnais, 27).

I want to emphasize the Oulipo has had from its start the two notions of critical reading and critical making in mind. Through literary endeavor the Oulipo seeks to “research” through analysis  and synthesis, the bringing together to develop new possibilities. The same attention to procedure and constraint allows this group to do both.

It is the particular procedural methods of bot-building and data-culling that place bots within an aesthetic tradition of making as philosophical experiment that is long-standing. The reason for this is not, however, to make a claim that we might categorize bots as literary or aesthetic artifacts, but, rather, to contribute to a dialogue about the ethos of making that might benefit those turning to critical compositions (procedural or not) as components of their scholarship.

Bot Making as a Critical Poetics of Inquiry, or, a Note Toward “Results”

Permit me for a moment a brief anecdotal interlude. My work and some element of my future research in rhetoric and composition might be located somewhere between the second and third levels. The bot allowed me to author a procedure and let some other process (that is, a process other than my own subjective creation) to generate new language. As a poet (as is evidenced by the other entries on my blog), I’m always seeking ways of constraining or enabling processes of language production, especially through (re)combinatory methods. My current poetic project, a manipulated erasure of the book A Land by Jacquetta Hawkes, does not employ predetermined procedures for which words I will remove from the original text, but does employ constraints for my methods of manipulation; that is, I can remove a punctuation mark or capital letter from earlier on the page if I did not use them. These constraints, however, still have much to do with my own choice with selecting language rather than the authoring of language procedures that are pre-imposed. My procedures have to do with a desire to honor the sentence as an important unit of construction, where the Oulipo’s formal constrained procedures rely on processes that can be replicated exactly.

I say this not to recommend some hybrid practice of traditional composition and constraint (though this is what I’m doing, and this also has a lot to do with my interest in genre, as the intersection of both “traditionally” authored language and constraints both self-induced and socially constructed), but to suggest that part of the motivation for this project was to undertake a more “purely” algorithmic practice as a mode of language generation and release of my anxieties surrounding singular authorship.

That twitter makes us compose in 140 characters necessitates efficiency with our insights and emotions, depending on our purposes for tweeting. Mark Sample himself reflects in a  Media Commons piece “On Sharing and Losing an Online Persona,” where he relates a Twitter adventure: being shut down for a fictional plane disaster “live-blog” via Tweet,  and discovering that there were parody accounts of appropriating his avatar and pretending to be him that people mistook for actually him. Ultimately there’s this: “That’s what sharing online gets you. Share enough of your personality and you’ve given the world a blueprint to be you. And you’re not even that interesting.” While it makes sense to those of us who value literary-ness in our texts to find bots that cull from literary or already published works as appealing from the view of aesthetics, Twitter is a living archive of material language moving with and across users in time and space. The language the everyday, even that which is “not even that interesting” when recombined into new formations reveals something of our human nature and the representation of our senses in this public sphere.

This bot was built to capitalize on those representations as much as to relinquish my own authorial agency. I authored the procedure and set it in motion. To be fair, however, I only sort of authored this bot. I culled code from a few different bots and asked for the assistance for some of the programming bits I couldn’t figure out, so I really consider the procedure collaboratively authored (though aren’t all procedures produced in this way?) Now I just get to take part as a critical reader of its outputs and its mechanism. Whatever rhetorical significance the actual tweets might have depends on the interaction of the reader and the ways in which reading slightly misaligned sentence constructions works in the body. While this bot is by no means a perfect program, its glitches are also what make the reader complicit in its meaning making. Do you take that word “undefined” as something a human tweeted or as a machine output? That depends, I suppose, on the reader.

To use Le Lionnais’s terms, I see this project as both analysis and synthesis. I am investigating, by recombination, language use in the Twitter-sphere, and also developing new possibilities for language construction through the algorithm of my bot. I recommend, therefore, an Oulipian orientation toward the Bot not to recommend that the Bot is necessarily generating poetry, but to suggest that we might learn something about creative computing and an ethos of making as well as hone our critical faculties by relying on the ideological foundations put into place by the poetic avant-garde.

Notes & Acknowledgements

NOTES: [1] This bot was created as the final project for a Materialities of Writing seminar at the University of Pittsburgh, taught by Annette Vee. [2] During the semester we made and/or did things and blogged about them, and the FlyOnTheWallBot 1.0 was created as one of those weekly ‘signments. [3] Given that I’d build a digital thing it seemed only appropriate to “host” my theoretical and makerly exigencies on this blog. Plus I wanted to link to a bunch of things. Plus I wanted a method of promoting my bot as a part of my scholarly/markerly endeavors. This, I hope, makes sense. [4] The bot was written in Google Script and hosted in Google Drive. If you would like to see the code feel free to contact me at moriahlpurdy [at] gmail [dot] com or via Twitter at @moriahlpurdy. [4] This used to be a longer post but I’ve culled it down for the sake of brevity and because some of the ideas in the rationale for the project felt like they were getting somewhere, but were not quite it (because, graduate school). What is left feels about right.
ACKNOWLEDGEMENTS: [1] Many thanks to my Materialities-mates, without whose weekly conversations about writing materials this thinking could not exist. [2] Many thanks to Tim Kirchner, whose expertise, patience, and encouragement helped to make the bot I’d conceived of in my mind a reality.