The Electronic Literature Organization’s annual conference was last week in Milwaukee. I hated to miss it, but I hated even more the idea of missing my kids’ last days of school here in Madrid, where we’ve been since January.
If I had been at the ELO conference, I’d have no doubt talked about bots. I thought I already said everything I had to say about these small autonomous programs that generate text and images on social media, but like a bot, I just can’t stop.
Here, then, is one more modest attempt to theorize bots—and by extension other forms of computational media. The tl;dr version is that there are two archetypes of bots: closed bots and green bots. And each of these archetypes comes with an array of associated characteristics that deepen our understanding of digital media.
Tully Hansen has already drafted a tentative taxonomy of Twitter bots, complete with a stunningly complete family tree. Tully presented a revised version of this family tree at the ELO conference on Thursday. Both versions of the taxonomy show a foundational split between two kinds of bots: independent bots and dependent bots. The former are bots that run independently of any other input or feedback and the latter are bots that rely on external input, often coming from Twitter itself.
It’s a very generative taxonomy. Every time I trace the different generic paths on the chart, I ask new questions about bots and how we think about them. The great divide between independent and dependent bots is especially useful. Lately I’ve been wondering how to reframe this difference in a way that focuses on structural qualities rather than operational functionality. This is where closed bots and green bots come in.
Closed Bots and Green Bots
My approach is inspired by Paul Edwards’ remarkable book about computing and the Cold War, The Closed World.1 Edwards reworks the theories of two Shakespearean scholars—Northrop Frye and Sherman Hawkins—in order to analyze the ideology of the Cold War. From Frye’s Anatomy of Criticism comes the notion of a green world (which typifies many of Shakespeare’s comedies, e.g. A Midsummer’s Night Dream) and from Hawkins comes the idea of a closed world (which typifies some of Shakespeare’s comedies and many of his tragedies, e.g. Hamlet). As Edwards puts it:
With its focus on containment and invasion, you can see why Edwards choose the closed world as a metaphor for Cold War ideology. The green world transcends authority, convention, and technology. The closed world basks in these things.
Long before Edwards borrowed the idea of a closed world to describe the mindset that drove Cold War strategy, Hawkins suggested that the closed world and green world map onto two enduring narrative archetypes: the siege and the journey.
Confused? Here’s a chart:
|Closed World||Green World|
|Archetypal Narrative||The Iliad||The Odyssey|
|Archetypal Location||Castle or Fortress||Forest or Meadow|
|Archetypal Movement||Unity of Place||Multiple Borders Crossed|
|Archetypal Knowledge||Science and Technology||Magic and Mystery|
|Archetypal Resolution||Escape or Invasion||Restoration of Community|
I don’t know whether Frye and Hawkins’ theories hold sway over contemporary medieval and Renaissance scholars. They are undoubtedly unfashionable figures to use to think about digital culture. It’s conceivable that Frye’s generic structuralism might be hip in a retro kind of way, but most literary scholars have probably never heard of Hawkins. Certainly I hadn’t, until I read Edwards. Yet, there’s something refreshing about the generic models Frye and Hawkins provide. As Hawkins himself put it in his 1967 article about the closed world in Shakespeare, “The green and closed world patterns provide a basis for comparison, not a rule of conformity…They permit us to draw together for comparison motifs, structural outlines, entire plays of different periods.”2 Hawkins invites readers to extend these archetypes to other narrative forms, not in order to classify literary works, but in order to draw out thematic nuances we might otherwise overlook.
Whether or not these archetypes have traction with literary scholars, it’s clear that the siege and the journey continue to resonant in popular culture. Take videogames, for example. Space Invaders (1978) depicts a siege, The Legend of Zelda (1986) is a journey. Little has changed in the past thirty years. Journeys dominate contemporary videogames (Skyrim, I am Alive), but the siege—especially in the form of an invasion—is common too. Think Halo, Quake, and many other first person shooters.
But, getting back to bots, I want to suggest that the closed world and green world are not merely thematic archetypes that apply to narrative forms. The closed world and green world are also archetypes for the generative processes of computational media. They are archetypes of procedural composition. Where does a procedural work—a rules-driven work—get its source material? From within itself, or from beyond itself? Is there what Hawkins calls a “unity of place” in the work, or does the work come about through transgression across thresholds and barriers? Is the work closed or green?
The recently concluded @everyword (Allison Parrish, 2007) and similar completionist bots such as @everyunicode (Ramsey Nasser, 2013) are archetypal closed bots. They proceed relentlessly through a list. They exist in a world all their own, a kind of procedural fortress. There’s something brutally self-referential, self-enclosing, self-disciplined about them. Closed bots always circle around to themselves. There’s no way out. Bots with predetermined templates like @metaphorminute (Darius Kazemi, 2012) or @streetsnsheets (Casey Kolderup, 2014) might have vast vocabularies at their disposal, but they nonetheless are closed as well. Given enough time, they will exhaust every possible combination of material. Closed bots seduce us with linguistic entrapment, as if language itself were under siege.
In contrast, green bots cross boundaries, absorb diverse material, and often surpass the limits of rationality. Bots that make use of other tweets are the most common green bots. @AndNowImagine (Ivy Baumgarten, 2013) and Adam Parrish’s @oneiropoiesis (2014) are stand-out examples. The millions of tweets out there these bots pull from are the faerie-enchanted woodlands of a Renaissance comedy. Anything seems possible. Established orders are temporarily suspended, and the world is turned upside down. That’s what happens when Walt Whitman channels an American teenager (or is it an American teenager channeling Walt Whitman?).
Variations of the Archetypes
At first glance, the closed/green archetypes appear to map onto Tully Hansen’s taxonomy of bots: closed bots seem like independent bots, while green world bots recall dependent bots. However, there are key differences between Tully’s functionality model and my archetypal model. There is a strict binary opposition between independent and dependent operations, while archetypes allow for variation. Indeed, our level of engagement in a generic work often corresponds proportionately to how much it deviates from the archetype we expect it to be. In other words, it’s possible for a bot to alternate between a closed world and a green world and points between—and it will likely be all the more interesting for it. My @_lostbuoy_ (2014) illustrates this dynamic, drawing from real-time oceanic and atmospheric data (green world) while mashing up the text from Melville’s Moby-Dick (closed world).3
Every archetype comes with a cluster of attendant characteristics. These are characteristics likely to be associated with an archetype but aren’t strictly necessary, at least not all of them at once. An attendant characteristic of the journey archetype is the appearance of a guide who provides help along the way. An attendant characteristic of the siege archetype will likely be a betrayal. If we shift from narrative archetypes to procedural archetypes, we can similarly find attendant characteristics. Closed computational media are likely to have a unity of voice—e.g. @notallbots (thricedotted, 2014) or @rom_txt (Zach Whalen, 2013). Green world computational media are likely—as in green world narratives—to involve cases of mistaken identity, so that it’s unclear whose voice is speaking. Mashup bots fit the bill here, such as @savehumanities (2013). Another attendant characteristic of closed world bots is that they are what Chris Crawford would call data-intensive, working with large datasets of images or text. Green world bots are more likely to be process-intensive, heavily dependent upon algorithms, natural language processing, and text manipulation.4
This model of procedural archetypes extends beyond bots to include other computational works. Consider two procedurally generated novels that emerged from National Novel Generation Month (a take-off of National Novel Writing Month): Nick Montfort’s World Clock and Darius Kazemi’s Teens Wander around a House. World Clock is fully self-contained; the Python source code provides a controlled vocabulary and grammar. With a paragraph for every minute of every hour across a 24-hour period, it is a vast work. It is also a closed world work. Meanwhile Teens Wander around a House is built from Twitter subtweets and dream reports from the DreamBank. It’s a work of assimilation, blending different domains of knowledge. It’s interesting how the structural origin of these two novels (closed/green) leak into their thematic concerns. World Clock focuses relentlessly on time, which is a stand-in for the empiricism and cold rationality of the closed world. Teens Wander around a House is punctuated by the dreams of children, dream worlds being the par excellence green world.
Closed is not Closed and Green is not Open
This framework of closed and green procedural media is not meant to be prescriptive. There’s no value in proscribing what procedural models creative coders should or should not experiment with. The archetypes are meant to be descriptive, and to give us traction as we think more deeply about the algorithms that delight us and intrude upon us. One drawback to this descriptive model are the terms themselves. “Closed” has become a bad word and it’s easy to see it as the negative term in the closed/green pair. Yet some of the most enduring works of literature demonstrate that closed worlds are powerful, entrancing, and worth thinking through. It’s important to remember that closed worlds can be richly textured and full of surprises.
Similarly, “green” is easy to misunderstand. Given the aura surrounding “open” in Internet culture (open source, open access, etc.), it’s tempting to retcon Northrop Frye’s “green world” into “open world.” But open is not the opposite of closed here. Green conveys something beyond openness. In the 21st century the green spaces—the spaces where rationality is breached and magic erupts—are no longer forests or glens. They’re subreddits and hashtags, SnapChats and Yik Yaks. That’s where the fantastical is these days—or at least, that’s where the irrefutable logic of the rational world recedes from view.
Closed and green. Two archetypes for the generation of novelty in the digital world.
Edwards, Paul. The Closed World: Computers and the Politics of Discourse in Cold War America. Cambridge, MA: MIT Press, 1996.↩
Hawkins, Sherman. “The Two Worlds of Shakespearean Comedy.” Shakespeare Studies 3 (1967): 62-80.↩
Moby-Dick itself is an intriguing blend of closed and green world narratives. The hunt is an epic quest across unbounded nature (green world) yet Ahab’s totalizing pursuit of the white whale resembles a closed world obsession.↩
Crawford is speaking solely of videogames, but others have extended the concept of data-intensity and process-intensity to other forms of digital media. See Crawford, Chris. Chris Crawford on Game Design. New Riders, 2003; Wardrip-Fruin, Noah. “Five Elements of Digital Literature.” Reading Moving Letters: Digital Literature in Research and Teaching. Ed. Roberto Simanowski, Jörgen Schäfer, and Peter Gendolla. New Brunswick, NJ: Transaction Publishers, 2010. 29–57.↩