Fictional what-ifs come in all shapes and sizes. Many are ephemeral brain farts that are as disposable as they are whimsical. What if Neo had chosen the blue pill? What if plants could talk? What if animals wanted us to eat them? What if we needed stamps to send email? What if Steve Jobs was an alien who merely returned to his home planet? What if Donald Trump turned the Whitehouse into a casino? Other what-ifs, born of exactly the same slice’n’ dice attitude to reality, turn out to be made of weightier stuff and provoke fascinating debates. What if animals brought a class action suit against the human race? What if the ancient Romans had invented the atom bomb? What if the USSR had won the cold war? What if the Axis powers had won World War II and divided up the United States among themselves? This last what-if is the provocative basis for Philip K. Dick’s Hugo Award–winning novel, The Man in the High Castle.1 Written in 1962 and set in the same year, the book explores an alternate reality in which the allies lost World War II after the Nazis dropped the first atomic bomb on Washington, DC. The political counterfactuals of the novel concern the machinations of Nazi Germany, which now controls the Eastern states of America, and of Japan, which controls the Pacific states, with the Rocky Mountain states serving as a neutral buffer.
But Dick’s novel is, more than anything else, a philosophical inquiry into our tangled understanding of reality versus appearance, fact versus fiction, authenticity versus artifice, and fate versus chance. His fictional Japanese occupiers of the Pacific states exhibit a fetish-like desire for the collectible vestiges of the prewar United States, collecting anything from old bottle caps to weapons and furnishings. Skilled counterfeiters realize tremendous profits in this seller’s market, and Dick’s novel follows both a purveyor of high-price antiquities and a producer of high-quality forgeries who later branches out into the creation of original items of contemporary jewelry. Dick uses these complementary perspectives to explore the inherent value of an artifact. What does it mean to say that something is a “fake” or the product of artifice? Does an object that “has history in it,” such as an object tied to a famous person or a pivotal event in history, have more intrinsic value than a perfectly functional copy of the same thing? Does the origin of a thing, or the intent with which it was made, wholly determine its usefulness to others? This is a novel in which fakes of all kinds abound, from people who are not who they seem to be or who they say they are, to competing histories that never completely persuade. The characters of the novel may hold differing views on “historicity,” but each in his or her own way attempts to project external meaning onto objects and events to see their way to an “inner truth.” In many ways, the themes of Dick’s novel are as applicable to our appreciation of artifacts made by autonomous machines as they are to more conventional man-made artifacts, and perhaps all the more so because our machines are themselves a special kind of man-made object. We ask many of the same questions of their outputs and face many of the same doubts. In this final chapter, we shall find that many of the themes of Dick’s novel chime with the opportunities and the concerns raised by our bots.
Consider the desire for historicity, that is, our need to find a connection between an artifact and some external context that can give it relevance and meaning; the search for history is ultimately a search for story and a desire to frame an object within a satisfying narrative. Dick’s novel has a cynical manufacturer of forgeries make the case that historicity is a comforting story we tell ourselves and others. Holding up two identical-looking Zippo lighters, only one of which was owned by Franklin D. Roosevelt (who was assassinated prior to World War II in the time line of the novel) he argues, “One has historicity, a hell of a lot of it. As much as any object has ever had. And one has nothing. Can you feel it? … You can’t. You can’t tell which is which. There’s no ‘mystical plasmic presence,’ no ‘aura’ around it.” An imitation Colt revolver fills the novel’s antiquities dealer with shame after he has been duped into selling it, but the fake proves to be just as effective as the real thing at killing two black hats in a shoot-out redolent of America’s Wild West. Yet we shudder at knockoffs and fakes even when they are made to exacting standards because their narrative is one of deception in which we are the dupes. Our bots do not subscribe to this narrative because they wear their artifice on the sleeves. We humans knowingly follow bots because they are bots, and not for any of the reasons that we buy knockoff products even when we know they are fakes. For we do not seek a false economy and an unearned narrative when we buy into the artifice of a Twitterbot: that an interesting tweet that has merit on its own terms was crafted by a bot is the narrative we seek, and it is no fake. Actually, to be more specific this is the metanarrative of a Twitterbot. The theatrical narrative of a bot, which is to say the playful pretense of what the bot purports to be when it tweets, can be something else entirely. To the extent that this specific narrative is a fake, it is an ironic fake designed to be recognized and enjoyed as such.
This puts the pretenses of a Twitterbot on a different plane from those of human creators who might seek to weave convenient fictions around the genesis of their own outputs. Samuel Taylor Coleridge’s 1797 poem “Kubla Khan” runs only fifty-four lines and can seem like a teaser trailer for an epic work of the imagination that would never come.2 The poem is remembered so fondly, at least in part, because of a creation myth that Coleridge spun about its writing. It seems the poet had awoken from a vivid dream, which may or may not have been drug induced, and had set about rendering the world of his imaginings in equally vivid detail when he was interrupted by a “man from Porlock” calling on business. By the time the man was shooed away, the well of the poet’s rememberings had run dry, forever imbuing the remaining stub of his poem with a wistful air of “if only.” Intellectual products are no more immune to our desire for historicity as physical objects. In 2011 Imbibe magazine3 featured this tipple from the golden age of jazz: “Named after the Louisville, Kentucky hotel where it was first crafted in 1917, the Seelbach cocktail is a classic mix of bourbon, Cointreau and both Angostura and Peychaud’s bitters.” But the true story of its creation was revealed in 2016 in a New York Times article, “That Historic Cocktail? Turns Out It’s a Fake.” The drink’s creator, Adam Seger, had falsely claimed to have rediscovered the all-but-lost recipe for what was once supposedly the hotel’s signature tipple, and drinkers latched onto this new connection to the Jazz Age with enthusiasm.
According to the New York Times, Seger “came up with an elaborate origin story involving a couple from New Orleans who had honeymooned at the hotel in 1912. The man ordered a Manhattan, the woman a Champagne cocktail. The clumsy bartender, spilling the bubbly into the Manhattan, set the mess aside and made the drinks anew. But the accidental mélange got the barman thinking. Soon, the Seelbach cocktail was born.”4 So into the false historicity of the cocktail Seger wove strands of a classic narrative of creativity, the enlightening accident, though Seger’s tale is no more convincing than the classic advertisement for Reese’s Peanut Butter Cups, which shows a truck carrying chocolate crash into a truck carrying peanut butter (yet the latter, like the whimsical fictions of a Twitterbot, lacks any intention to deceive). If this all sounds like William Burroughs’s cutup method of creation applied to foodstuffs rather than textstuffs, it may be because so many of our creation myths, not unlike the heroic tales of Campbell, Vogler, and Propp, necessitate a daring crossing of boundaries that few have crossed before. Indeed, sometimes the crossing is so daring that only an act of “accidental mélange” or fateful chance can seem like an appropriate call to adventure for our hero. These creation myths and origin stories work because we believe them, more or less, and this places them into sharp contrast with the framing narratives of our Twitterbots, which work precisely because we don’t believe them. It is as though those bot narratives began with “Once upon a time” or “A bot walks into a bar,” so willing are we to suspend disbelief and join in the bot’s pretense. We accept the narrative for what it is: an invitation to enter a certain mind-set and play.
Consider the framing narrative of Nora Reed’s @thinkpiecebot, which doesn’t actually produce op-ed think pieces but does generate provocative titles for the silliest of what-if articles. As stated in the bot’s FAQ, Reed abhors the fashion among the journalistic great and the good to diagnose the excesses of the millennial generation.5 Perhaps it is the perceived smugness of millennials, with their preference for craft/organic/free-range/gluten-free products and their dismissal of conventional consumerism, that fuels the industry of patronizing articles that shout “Hipsters have ruined [X]” or “It’s the Fault of [Group] that [Z].” Whatever the deeper motivation for these patronizing articles, Reed designed @thinkpiecebot to push back, not with scorn but with a gentle, generative satire. The bot was created using Cheap Bots Done Quick and Tracery, and is simply but effectively a randomized filler of templates that capture the signature syntactic and lexical preferences of op-ed headline writers. Examples include “Has Toast Hit the Tipping Point?”, “Hipsters Are Killing the Hipsters Industry,” and the wonderfully provocative, “As a White Dude, I Really Think We Should Focus on Indie Comics Instead of the Global Rise of White Supremacy.” Templates that take two or more randomized fillers (which come from a central casting agency of millennial subgroups and Zeitgeisty topics) engineer a crossing of boundaries that is very much in the spirit of Gysin and Burroughs’s cutups. And as with even the best of Burroughs’s own cutups, this generative backstory gives its contents a greater hold on our imaginations than they might otherwise command. Reed’s bot gently mocks what it pretends to be, but its satire is honed by the simplicity of its approach: if a bot can generate pseudorealistic headlines with so little effort, we can hear the bot ask: “How much imagination goes into writing the real thing?” And Reed’s bot is the answer to its own tacit question: “Not so much.”
A bot’s narrative backstory can create distance, as is often the case with satire, or it can foster empathy, even stirring our emotions for nonsentient things. Kate Compton (@GalaxyKate)’s bot @LostTesla manages to achieve mainly the latter, with just the merest hint of the former. (As you might expect, Compton’s bot also uses Tracery and CBDQ, and a link to her source code6 can be found in the bot’s Twitter bio.) At a time when speculation is rife about the coming world of autonomous cars and when investors have eagerly funded Elon Musk’s electric, and soon to be driverless, Tesla, the @LostTesla bot imagines what it would be like for a sentient vehicle to miss its human owner. Think of Homer’s Odyssey told from the perspective of a Tesla trying to make its way home to the garage of a waiting Penelope, and you’ll have some idea of the wistful tenor of @LostTesla’s tweets. Onto this backstory Compton has layered the personality and curiosity of a child, so that the bot seems to take joy from the sighting of a squirrel or show concern over its appearance (because Mommy will be angry?) or promise to flip its internal settings to best behavior. So its tweets run from the sweet, such as, “it is raining i will be a good car,” to the yearning, such as, “I watch my reflection in my sideview mirror. i’m .. muddy / i watch some cows. many chickens. / can i be a chicken with you?” and, “hello red car will you drive with me / hello gold car let’s drive together” to the Zen-like, “there is a pigeon on the right / i will remember / i am a being of hopes and dreams.” The bot represents a masterful example of what a thoughtful bot builder can achieve just using Tracery or CBDQ, and to read through @LostTesla’s tweets is to be reminded of HAL 9000 singing “Daisy, Daisy” in 2001: A Space Odyssey. Though we recognize its tweets to be the stuff of simple, systematized whimsy, the bot can certainly make us think, and perhaps feel too. As we imbue our artifacts with ever greater intelligence and a semblance of sentience so that they may do an even better job of serving our needs, what are our responsibilities to these emotionally resonant objects? When it comes to the creation of evocative what-ifs, this really is the best of bot worlds.
The best bot narratives are pretenseful but unpretentious, fake but not phony. The philosopher Harry Frankfurt took pains to sketch a theory of the concept of bullshit as it is used in everyday language and concluded that bullshit is no ordinary act of lying or deception but a use of apparent “facts” without a due regard for the truth of a situation.7 We bullshit not when we know that what we are saying is wrong but when we do not care if it is right. In this sense of the word, Twitter is home to a great deal of “bullshit,” perhaps the most egregious examples of which concern a recent trend for fake “news.” Facebook, Twitter, and other social media allow the most preposterous claims to be spread by those who care very little, if at all, about truth. Sex dungeons in pizzerias run by your political opponents? Check. Millions of illegal voters tipped the popular vote in your opponent’s favor? Check. Sharia law to be declared in a Texas town? Check. Fortunately, while bots might well be used to automate the spread of this kind of corrosive nonsense, most bot-framing strategies score very low marks on the bullshit-o-meter, and as shown by Reed’s @thinkpiecebot, our bots can use their own brand of fakeness to expose the bullshit of others without succumbing to bullshit themselves.
Bot builder Mark Sample has repurposed a polemical claim by 1960s singer Phil Ochs—that “a protest song is a song that’s so specific you can’t mistake it for bullshit”—to produce a comparable rallying cry for bots: “A protest bot is a bot that’s so specific that you can’t mistake it for bullshit.” Both Ochs and Sample use the term “bullshit” is a way that gels with Frankfurt’s definition, as specificity in each case demands a moral clarity that has no substitute in angst-ridden pabulum or in catchy nostrums. For Sample, a bot that invents things that are simply not true can still be a “bot of conviction” if it invents them for the right reasons.8 If you wonder how a bot that invents wholly counterfactual scenarios might show due regard for the truth, Sample poses an interesting solution: make your bots data driven, so they may build up a picture of a larger truth that is cumulatively accurate at the statistical level while using constructive pretense (as opposed to pure bullshit) at the level of individual tweets. Sample’s @NRA_tally is such a bot of conviction that uses counterfactualism rather than bullshit to invent Twitter reports of spree shootings, which it pairs with stock comments from the National Rifle Association (NRA) to highlight the gulf between opposing sides of the gun control debate. While it thankfully fabricates its shootings (the bot tweets a gun-related atrocity every four hours), @NRA_tally draws its data for the number and type of victims, the type of location, and the model of weapon from the statistical record, and pulls its NRA responses from real NRA commentary. Indeed, one can credibly argue that the bot shows more respect for the reality of gun crime in America than certain news organizations with a partisan belief system to uphold.
Frankfurt’s analysis is predicated on the power of “bullshit” as a pejorative label, but what qualities are the opposite of “bullshit” and how might we build our bots to embody those qualities? Dick’s novel considers a pair of designations for human craftsmanship that he borrows from Eastern philosophy. The first is wabi (or wabi-sabi), a Japanese term that loosely translates as lean, spare, and graceful. An object has wabi if it lacks unnecessary frills and fulfills its function with a no-fuss, no-bullshit grace. An artifact with wabi will have earned its imperfections and will wear them well as a sign of its historicity. Due to its restriction on the length of tweets, Twitter seems an ideal place to look for the linguistic equivalent of wabi: a well-crafted news headline may have wabi in spades, as might a finely wrought joke that says no more than it absolutely must to achieve its humorous effect. The “novels in three lines” of modernist writer Félix Fénéon, whom we met in chapter 1, appear to have been constructed with wabi as their chief artistic motivation. Wabi abounds whenever Twitter is used masterfully as a medium, as in the @novelsin3lines account that was retrospectively created to showcase Fénéon’s oeuvre in English, yet because the designation is a term of discernment, wabi is, sadly, far from the norm. The Japanese in Dick’s novel are connoisseurs of wabi and eagerly seek it out in items of collectible Americana, but they cast a cold eye on anything that is new or lacking in history, or seemingly without a useful function that might offer a larger context in which to judge its wabi-ness. A forger of antiquities in the novel turns his hand to making artisan jewelry, and a dealer in antiquities takes some of the jewelry on contingency. The dealer, Robert Childan, presents one of the pieces, a decorative pin, to Paul, a valued Japanese customer, but he is not impressed with the offering. At first Paul is confused: the object, a shiny “blob” of polished metal, seems altogether risible, and his friends snigger in agreement. Though Paul is embarrassed for the dealer, he cannot stop himself from stealing glances at the pin, for reasons he does not yet understand. The object haunts his thoughts, and when Paul meets again with Childan, he offers these observations:
It does not have wabi, Paul said, nor could it ever. But—He touched the pin with his nail. Robert, this object has wu.
I believe you are right, Childan said, trying to recall what wu was; it was not a Japanese word—it was Chinese. Wisdom, he decided. Or comprehension. Anyhow, it was highly good.
The hands of the artificer, Paul said, had wu, and allowed that wu to flow into this piece. Possibly he himself knows only that this piece satisfies. It is complete, Robert. By contemplating it, we gain more wu ourselves.
We do not need a larger frame of reference in which to appreciate wu: an object with wu is sufficient onto itself, exhibiting inner balance and harmony. Broadly speaking, whereas wabi is a quality found in man-made artifacts that satisfy their functional demands with unshowy elegance and grace, wu is a quality most often found in natural objects that have no designer or no predetermined function to serve. We might perceive wu in the symmetry of an image woven from a complex cellular automaton, though it is stretching the point somewhat to talk of wabi and wu as possibly inhering in linguistic tweets, especially in the tweets of an automated bot. Nonetheless, we gain a certain amount of leeway by having the term bullshit anchor the other end of our aesthetic spectrum, and while tweets hardly count as natural objects, it is no accident that we use the word sublime to describe both the ineffable wonders of nature and the wonders of a poetic turn of phrase. The most exquisitely wrought aphorism, for instance, combines a lightness of touch with the sense that one could not have said it better, as not a single word can be profitably changed. Such a phrase needs no historic frame of reference in which to be appreciated, save for the frames that unite us all: the frame of language and the frame of human existence. So the bon mots of Oscar Wilde and Dorothy Parker have inner harmony, complementarity, and balance in abundance, and to enjoy them is to feel that some wu-like quality has flowed straight from the writer into his or her words. Aphorisms such as these are self-contained and complete, and as Dick’s character Paul suggests, by contemplating them, we gain more of their wu-like quality for ourselves.
Fortunately, wabi and wu are not all-or-nothing concepts; rather, they are a continuum along which our bots might gradually progress with time. Just as @LostTesla’s tweets are occasionally Zen-like, there is a certain wu-like self-sufficiency in the solipsistic metaphors of @metaphorminute, which showcases the exuberance of language without trying to mean anything at all, or in the way that @NRA_tally’s tweets counterbalance the actions of two different kinds of gun fanatic. To the extent that the tweets of any Twitterbot exhibit either wabi- or wu-like qualities, it is because the bot has been designed to embody those qualities, so that they might flow from builder to bot to tweet. The bot itself may thus exhibit wabi if it embodies a simple but appealing idea with leanness, spareness, and grace. As such, the possibility does exist for our bots to add to the collective wabi and wu of the Twittersphere with the products of their linguistic and visual invention.
The most intriguing what-if in Dick’s novel The Man in the High Castle is not his alternate history of an allied defeat in World War II: that conceit is as evergreen as the notion of using a time machine to assassinate Hitler or warn FDR about a Japanese attack on Pearl Harbor. No, the most counterintuitive counterfactual is Dick’s suggestion that the people of his alternative time line, living under the totalitarian yoke of the Third Reich and Imperial Japan, would make the I Ching, the ancient Chinese system of oracular divination, an integral component of their everyday lives. Dick’s characters consult the texts of the I Ching for insight into all of their moral questions, big or small, and it comes as naturally to them as tossing a coin, playing paper-rock-scissors, or reciting “eeny-meeny-miny-moe.” But rather than give decision makers a binary random variable (whether heads/tails, win/lose, or it/not-it), the I Ching serves up a random signpost into a decision space of sixty-four possibilities, or hexagrams. Dick’s characters generate the hexagrams of the I Ching, blocks of six lines apiece where each line is either solid or broken, giving 26 = 64 possibilities, by throwing yarrow stalks or by tossing coins. They then look up an analysis for the hexagram that chance has given them in volumes of ancient commentaries, in a process called “consulting the Oracle.” To understand how Dick’s characters use the randomness of the I Ching to systematically weave a meaningful narrative around their actions is to understand how chance can be purposely harnessed by any decision-making agent, whether a human or a bot.
We all find ourselves blocked and stuck in a creative rut from time to time, and it is more easily said than done to look at our problem with fresh eyes. One way to force a new perspective upon ourselves—according to creativity mavens such as Edward de Bono– is to actively engage with a random but meaningful stimulus.9 We might, for instance, open the dictionary at a random page, pick a word with our eyes closed, and then try to integrate one or more senses of this word into our thinking about our problem. Though utterly out of left field, these fresh elements may be just the stimuli we need to escape our rut. But we are not limited to the dictionary when we play this game, for we could just as easily use the Bible, the Quran, the Torah, the Guinness Book of Records, Bartlett’s Familiar Quotations, Wikipedia, or the tweets of Donald Trump as our source of external stimulation. This simple method has a surprising provenance and is a whimsical update of an ancient practice called bibliomancy in which, for example, Christians looking for moral guidance might pick a seemingly random chapter and verse from the Bible in the hope that God, or providence, has guided the selection, just as a Muslim might do the same with the Quran. Of course, any randomly chosen text fragment is not an answer in itself, but if one believes that the selection has a divine mandate, then one will look all the harder to see its potential relevance. So when we strip away the veneer of mystical mumbo-jumbo from the I Ching and set aside the notion that it allows us to read “the tenor of the universe” at a given moment in time (as one of Dick’s characters memorably puts it), what is left is an ancient version of de Bono’s dictionary method of bibliomantic inspiration, albeit one that has attracted volumes of sage commentary from ancient philosophers. This is the real psychological value of the I Ching: it cleverly exploits randomness in a process of systematic self-examination. Its random stimuli may come from without, but the answers to our specific questions must still come from within.
Dick was turned on to the I Ching in 1961, a year before he wrote The Man in the High Castle, and by all accounts he took it rather seriously as a method of inspired decision making. Just as his characters frequently “consult the Oracle,” and choose their actions to fit the hexagrams that they randomly generate, Dick generated the hexagrams for them at these junctures not by inventing what his plot required but by obeying the I Ching himself. That is, he would throw his own yarrow stalks to form his own hexagrams, which would then dictate significant aspects of the plot when they were integrated into the text as character actions. This does not seem so very different from how one might write a sword and sorcery novel by co-opting the dice-based mechanics of Dungeons & Dragons to choose among plot outcomes, and we can think of Dick as a high-brow dungeon master. 10 It is a testament to his discipline as a writer that his random D64 rolls sometimes took his story down avenues that Dick would have preferred not to pursue and cut off others that might have better reflected his desired shape for the story. Yet his use of randomness was not deterministic, as it is in simple D&D bots, because the I Ching is not deterministic: it uses randomness to engage subjective thought processes, not to determine the results of those processes. Nonetheless, by using randomness systematically, with a disciplined and almost algorithmic respect for the results of stochastic processes, Dick used an approach to plotting that is not so very different from the algorithmic storytelling of our what-if machines. A bot can likewise use randomness as a guide to decision making without being wholly determined in its actions by the results of random number generation. We have seen, for instance, how random outcomes might decide the high-level structure of a plot by determining the next triple of actions in a story arc, and how a bot may yet control how each of these actions is to be fleshed out using knowledge of the characters concerned, perhaps with a bespoke piece of creative dialogue. The key to balancing randomness and creative action is not to overdetermine the link from random stimuli to concrete outputs, but instead to view randomness as a high-level means of picking among different modes of low-level engagement.
To imagine what might happen were a writer to surrender even more control to a stochastic system such as the I Ching or D&D (or even Scéalextric) we need only look to Dick’s novel within a novel. The “man in the high castle” of the title is a character named Hawthorne Abendsen, the enigmatic writer of a novel titled The Grasshopper Lies Heavy that offers an alternate history of World War II in which, shockingly, it was the Allies that won the war, turning Germany and Japan into client states. This alternate-alternate history serves as a beacon of hope for the people of the postwar United States, making Abendsen, its controversial writer, such a high-profile target that he is said to live in a fortress named the “high castle” in the Rocky Mountain zone. Abendsen’s history is deeply at odds with the time line of Dick’s novel, but it is also strangely different from the history of World War II as we all know it. The United States prevailed at Pearl Harbor because of the foresight of President Rex Tugwell, succeeding an FDR who, like Tiny Tim, did not die after all. It is Tugwell who ensures that the US fleet is not in port during the Japanese attack. Abendsen foresees the Allies falling to rancor among themselves after the war, with Britain winning a new cold war with the United States. However, he is not forthcoming when pressed on how he comes by his book’s revelations, leaving it to his wife to admit his debt to the I Ching: “One by one Hawth made the choices. Thousands of them. By means of the lines. Historic period. Subject. Characters. Plot. It took years.” She paints her husband as little more than the CPU that executed the plot-deciding algorithm of the I Ching to tell a tale that is at once both his and not his. Dick is being ironic, of course, as he gently mocks his own reliance on the I Ching, but his larger point is that history is just another story in which we are all mere “characters.” Writers can tell such compelling stories using the mechanisms of simple chance because our own lives are subject to the very same mechanisms.
Users of the I Ching “consult the Oracle” with a specific question in mind. For instance, one of Dick’s characters poses the question of whether Abendsen’s novel is fiction or genuine history and generates the six lines of hexagram 61, Chung Fu (“Inner Truth”), with her coin tosses. With this, she infers that the novel is indeed factual and happily concludes that it is her own world that is fictional. But imagine a Jeopardy!-like version of bibliomancy, in which users generate the answer first and then find the question that fits this answer. In fact we all do this, insofar as reading makes bibliomancers of us all. Every time we read a book, or a news article, or a tweet, we cannot help but bring our own life experience to bear, to view the actions of another person or an imaginary figure as though they might be informed by, and inform in turn, our own issues and goals. Each new status update that pops into our Twitter time line invites us to see a relevance to our own lives. Although the I Ching would be a great topic for a Twitterbot—imagine a bot that delivers hexagrams and commentaries in response to any user who tweets the hashtag #iChingMe, using a sentiment analysis of recent tweets to guide its process of “oracular divination”—our bots already offer nuggets of thought-provoking text from a dynamic book that may be as specific as its own knowledge base or as general as the web or indeed all that language will allow. Consider the outputs of Parrish’s @everyword, which tweeted every word of the English language in alphabetical sequence. Had users viewed its outputs as mere word listings, it would never have garnered the bulk of its seventy thousand followers. Even if many of its frequent lexical intrusions into our timeliness were ignored—and who can say that every word in the dictionary is worth tweeting?—it only takes a few percent of its outputs to attract our interest and stir our thoughts for such a bot to make a small but meaningful contribution to our day. These bots may not be able to read or distill “the tenor of a moment in the universe,” but in their random exploration of the space of all possible signifiers for such moments, they create a potential for synchrony, in which a bot’s outputs may occasionally (if quite accidentally) capture the mood of the Zeitgeist. So when @everyword tweeted “woman” on May 14, 2014, in the same week that the New York Times fired its first female executive editor for supposedly being “too bossy,” the bot’s followers may have felt that it was providing the real reason for Jill Abramson’s dismissal. It requires a willing suspension of disbelief to think so, but it takes a comparable suspension of our most critical faculties to usefully engage with the I Ching too. This willing suspension is not willful ignorance, but a recognition of the value of nonliteral modes of expression and of nonliteral approaches to meaning.
The interviewer Charlie Rose offered this succinct analysis of the mainstream media’s failure to predict a Trump victory in the 2016 US presidential election: “Those [on the left] who took him literally did not take him seriously, while those [on the right] who took him seriously did not take him literally.” So the race was swayed less by those who believed Trump than by those who believed in him, with the businessman operating at, and thriving at, a level of nonliteralness that was unprecedented (or to use his own word, “unpresidented”) in modern times. 11 We can take a message seriously without taking it literally, and in fact the most serious analysis may be a nuanced nonliteral interpretation that goes to the nub of a speaker’s personality or worldview. Whether we are talking TwitterTrump or Twitterbot, @realDonaldTrump or @DeepDrumpf, it seems that seriousness and nonliteralness often go hand in hand: each wants others to take its tweets seriously enough to give them the charity of a figurative interpretation. In one instance, Trump described his own seemingly specific words as “an euphanism [sic],” acknowledging what many supporters had already intuitively grasped: this nonliteral pitchman uses language figuratively and hyperbolically, to convey not facts but aspirations, and not policies but attitudes. Fortunately, our bots have neither the power nor the privilege of shaping international trade and nuclear policy with their tweets, and so it is a different kind of seriousness we wish for them—the seriousness accorded to a comedian developing a playful but astute conceit (think Willy Wonka’s “a little nonsense now and then is relished by the wisest men”) rather than that which is accorded to an influential politician or statesman.12
Traditionally we knew—or seemed to think we knew—where we stood with politicians and comedians and the different brands of seriousness wielded by each. We are less certain now as the lines blur between the two and it becomes easier to parlay one kind of seriousness into the other. Just as in Dick’s novel, the truth of different narratives and time lines becomes that much harder to discern. Did Japan and Germany really win World War II, or is the truth closer to the alternate history outlined in Abendsen’s The Grasshopper Lies Heavy? Dick’s characters can only resort to the nonliteralness of the I Ching for answers. Did Trump really win the electoral college by a “landslide,” and would he have won the popular vote too if not for the actions of millions of illegal voters? Skeptics can look to the mainstream media for fact-checked analyses, but in an age where the authority of once-respected media outlets has drained away, facts carry all the weight of an I Ching divination to those who prefer to ignore what they do not want to hear. It seems that Sample’s observation that a protest bot is too specific to be mistaken for bullshit may no longer be quite as true as it once was, at least insofar as protest bots now operate on both sides of the political divide and are not inherently indebted to any one version of the truth. When the counterfactual is presented as factual and Frankfurt-style “bullshit” and Trump-style “euphanism [sic]” become valuable commodities, our bots will inevitably be used to fully realize this commoditization on an industrial scale.13
In language a nonliteral interpretation is often the only interpretation that can make any semantic sense, yet in art, nonliteralness is often a choice: we can take an artifact at face value or choose to look for its figurative potential to mean something more than what appears on the surface. It is not just high-brow art that affords this duality of interpretation and appreciation. A zombie movie is the ultimate two-level construct, a metaphor made flesh (oozing necrotic flesh to be exact). At the surface level, a zombie movie is as thrilling as a ride on a ghost train run by dubious carnies. Yet at a deeper level, if we care to look, we can find biting metaphors for rampant consumerism, political extremism, brain-rotting conformity, or implacable, unthinking rage of any kind. Zombies love brainsss, and as counterintuitive as it may seem, sometimes the place to look is inside the heads of the people who make these films. George A. Romero’s 1968 Night of the Living Dead was not the first zombie movie, but it was the first classic of the genre.14 Made during the Vietnam War and the struggle for civil rights, the film can be viewed as a metaphor for either or for both (tellingly, the black hero survives a rural onslaught of zombies only to be killed by racist thugs). Romero’s 1978 sequel, Dawn of the Dead, moves the action to a shopping mall where a group of survivors wait out the apocalypse in relative comfort, allowing the movie to poke fun at the mindless consumerism of the modern world. Romero’s zombies are shambling shufflers who are most threatening when they glom into hordes of slowly encroaching death, but other movies, such as Danny Boyle’s 2002 film 28 Days Later and Zack Snyder’s smart 2004 remake of Dawn of the Dead, turn the undead villains into frenzied runners with all the kinetic energy of rabid dogs. The metaphorical potential of these treatments finds fertile ground in the rapid spread of the zombie infection and in the ambiguity of who is and is not infected. A mainstay of every treatment is that there can be no reasoning with zombies: the best that one can do is to run, hide, or destroy, and suspected zombies must be isolated, quarantined, and delegitimized so they cannot compromise the safety or contaminate the purity of what remains of civil society. Political metaphors are never far from the surface in a what-if setup like this, so zombies have come to represent the dangerous spread of political dogmas rooted in anger and fear.
We might like to think of our Twitterbots as automated Willy Wonkas, each inviting its followers to “come with me and you’ll be in a world of pure imagination,” but it should surprise no one that bots can also be designed to automate the spread of our darkest and least whimsical ideas.15 The term zombie has long found favor in the online world as a metaphor for “infected” computers whose operation is subverted to perform mindless actions on behalf of a malevolent controller, but it is social media networks such as Twitter that enable the fullest expression of the dark potentialities of the computer-as-zombie metaphor. Just as in a horror movie, we may not know (even if we do suspect) that another user of the platform is a zombie or bot, not just because some bots are so well designed that they can fool others but because so many humans craft for themselves an irrational, implacable, one-track, context-insensitive persona on Twitter that—in the short term at least—makes them difficult to distinguish from bots. Twitter trolls are as sensitive to shibboleths and linguistic dog whistles as Watcher bots are to the hashtags or misspellings or turns of phrase that beckon them to action. They foist themselves and their views into online conversations with all the tact of a bot that is as eagle-eyed as it is flat-footed. Ironically, their goal is to inform and misinform, to spread information and disinformation, to seemingly bring a black-and-white clarity to a nuanced debate while muddying the waters as to the motives of those on the other side of an ideological divide. Just as they can seem programmed in their own rigidly unwavering viewpoints, they often accuse their opponents of being rigid automata themselves. In this fractious setting, it should not be surprising that the seemingly programmed views of one or both parties to a debate can be given programmatic form in the mechanics of a Twitterbot.
Some bots are designed to force themselves onto their targets, to call out what they see (or have been programmed to see) as egregious behavior by others. The @ImposterBuster bot of Yair Rosenberg and Neal Chandra is a troll hunter that, as its bio puts it, “exposes racist trolls impersonating Jews and other minorities.”16 The bot targets a specific subspecies of troll: those who try to pass as a member of a social group they despise, so that they might discredit the group from within. These trolls post extreme views from accounts with ethnic handles (in a rerun of medieval blood libels, some even pretend to be rabbis) with hashtags such as #blacklivesmatter, #ImWithHer, or #NeverTrump that mark them as members of the group they wish to taint with a strain of intolerance. Yet its pursuit of trolls turns the hunter into what it hunts, for though hunter and its prey may pursue different strategies and promote opposing strains of political thought, they can be equally zombie-like in their approach to intellectual debate on Twitter. Other political bots, such as @EveryTrumpDonor, are less hunter-seeker than namer-and-shamer. This bot repackages every financial contribution to Donald Trump’s presidential campaign as a first-person tweet that some will see as an admission of guilt. The frankness of each AA-like “confession,” which comes directly from Federal Election Commission filings, lends @EveryTrumpDonor a specificity that elevates it into a no-bullshit protest bot in the sense defined by Mark Sample.17 The bot does not hunt trolls directly but seeks to shame those who would feed or fund a troll. In contrast, other political bots use a Twitter equivalent of clickbait to actively feed trolls, in the belief that a troll who feeds on the offerings of a bot will have less bandwidth to prey on the offerings of human Twitter users. Bots such as @arguetron are generators of inflammatory twitbait that use templatized provocations to lure in, and tire out, those of the opposing political persuasion. Its templates include, “No one should ever listen to [X]” and “[Y] has a talent for saying things completely disconnected from reality,” while its Xs and Ys run from Julian Assange to Fox News, Gamergate, Pat Robertson, and Lena Dunham. Its emanations can seem apropos of nothing, with its tweets hanging like shriveled berries on a bare tree, yet the bot recycles the Twitter norms of strident political posturing so effectively that its unsuspecting targets come willingly to nibble.
Such bots may channel political abuse away from other humans, but they do little to reduce the overall anger and hostility of human-human interactions on Twitter and may even add fuel to the fires of discontent. A baited troll is rarely a sated troll, and the zombie-like workings of protest bots often achieve little more than a hardening of the battle lines between opposing camps. For while such bots show an ability to lure in an opponent, they exhibit very little capacity for changing that opponent’s views on a divisive topic. Indeed, once a bot’s voice is recognized for what it is, a cookie-cutter for political provocation, trolls are only encouraged to dismiss any human who uses the same linguistic norms to express the same or comparable views as being equally mechanical and template based. Bots like @arguetron may offer brainsss (or brain food) for once-human zombies to feast on, but they allow those zombies to characterize all of their opponents as frauds. Templates lend themselves to rigid patterns of thinking and speaking, but satire based on templates can cut both ways: a bot can use templates well, to suggest that others are slaves of an overly simple controlling idea, but in doing so, it may reveal that our own views of an opponent are just as rigid and unnuanced. When both sides use bots to present themselves or their opponents as inflexible automata with kneejerk responses to complex issues, no side can call itself the winner because each is dumbed down by its view of the other as unthinking zombies. The bots with the simplest and most provocative message may nonetheless grind out a victory on points, as the simplicity of templates gives them the edge over those whose messages rely on nuance, fact, and a willingness to see that complex problems often require complex solutions.
Twitter provides fertile ground for automated bluster and what has now come to be called computational propaganda by researchers who track the deleterious effects of bots on political discourse (the work of one such group of researchers can be found online at politicalbots.org). For instance, the use of hashtags can be as effective as a MAGA trucker hat in marking out the political leanings of a user, but just as hats cannot validate the true feelings of their wearers, at least outside the realm of Harry Potter and @sortingbot, hashtags are just as open to satirical use and abuse. Unfriendly agents may thus exploit and colonize the hashtags of their rivals, to insinuate themselves into their conversations under a false flag. In this way are proxy wars fought by our bots. Thus, supporters of Hillary Clinton in the 2016 presidential election cycle unleashed their bots to echo her battle cries and engage the supporters of her opponent, Donald Trump, on their territory, as marked out by their hashtags, and the supporters of Donald Trump did the same, bringing the battle to Hillary Clinton and her supporters via their hashtags. Hashtag “colonization”—say, when Clinton supporters use #MAGA for satirical effect or when Trump supporters use #ImWithHer to hurl brickbats at Clinton—devalues the hashtags of both sides, which is as good a reason as any for our bots not to use preexisting hashtags in their tweets unless they can bring something original and witty and openly bot generated to a conversation. Inevitably, when so many bots travel so widely on Twitter, bots of one political strain must often engage with those of another, like two zombies that shuffle around each other because each is unsure of the other’s capacity to provide brainsss. We might find comfort in the idea that a real zombie apocalypse must eventually run its course as the supply of human brains dwindles, but zombies on Twitter may live forever (if Twitter lives on) by continuously feasting on the droppings of other zombies.
So are there protest and propaganda bots that are not zombies, and if so, how might we tell the difference between zombie and nonzombie in a way that is not self-serving? We need more than a codification of the view that “our bots are not zombies because they are ours; your bots are zombies because they are yours.” We will find no hard criteria in which to anchor this ontological distinction, but we might as well begin with the idea of poetry. There is an urge toward creativity in the best of human protests, a reach for the figurative, the poetic, or the playful that shows the protester to be engaging at the level of ideas as well as utterances. As good a taxonomy as any for politically charged bots can be found in Leonard Cohen’s song “Bird on the Wire,” for our bots are our birds on our wire, flying in circles with as much or as little altitude and speed as we care to give them.18 Some, like @arguetron, are “like a worm on a hook” that baits a trap for hungry trolls, and others, like @NRA_tally, are “like a knight in an old-fashioned book,” embodying a certain notion of social value. Yet bots like @NRA_tally and @EveryTrumpDonor preach mostly to the converted, with their very existence on Twitter mattering more than any particular tweet they might generate. It seems fair to infer that most followers of @EveryTrumpDonor are less interested in a $200 donation by a Texas dentist to Trump’s presidential campaign than in the idea that any donor at all can be exposed to public scrutiny via the actions of a Twitterbot. These bots turn “following” into a political action, and the bots reward their followers with a sense of belonging and of having a tireless champion. Like the knight in Cohen’s song, these bots proclaim, “I have saved all my ribbons for thee.” Yet the most intriguing bots are also the least predictable, generating carefully packaged ideas that matter more than any single idea that the bot itself might embody. These bots strain against their guide wires, conveying at least the sense that they might occasionally transcend their limits and break free of our control. To repurpose Cohen’s words, such a bot is “like a drunk in a midnight choir,” but one that does more than screech another’s distinctive words as an off-key caterwaul. Drunks sing their lines zestily, with an unhinged and irreverent inventiveness, especially if they forget, or never knew, how a line is supposed to go. In this irreverence and inventiveness lies the all-important power to surprise. All things considered, we should prefer our bots to act more like creative drunks than unthinking zombies.
Some bots are designed to tweet every day of the year, while others will be as seasonal as eggnog, green beer, or Cadbury crème eggs. So a Twitterbot offering Halloween costume suggestions might run from September to the end of October each year, while one that suggests offbeat gift ideas might tweet only in the run-up to Christmas. Let’s suppose we set out to build a seasonal bot to poke fun at—or, more seriously, to protest—the official appointments of the new US president-elect, which in 2016 was Donald Trump. (If Hillary Clinton had won the election, our bot could just as easily take aim at her picks.) Such a bot will run for less than three months (November 9 to January 19) every four or eight years when a new president and a new administration takes power. Donald Trump’s own Twitter-lofted kite flying regarding his appointments made this a timely topic for a bot at the end of 2016, with the president-elect’s own tweets lending an unprecedented air of reality TV artifice to his putative picks. As speculation is just another form of invention that is partially grounded in fact, our bot can show a degree of creativity in individual tweets while parodying the selection process as a whole with its celebrity-obsessed modus operandi; that is, each specific tweet has an opportunity to make a reader think or laugh, or both, while the bot’s lax grip on reality—as shown, for example, by its willingness to nominate fictional or dead people to important government positions—can serve to satirize Trump’s real-life transition team’s understanding of both their task and the nature of government.
As a minimal zombie baseline, we can start with a set of templates of the form “Trump taps [X] for Secretary of [S],” where [X] is a randomly chosen proper name from a list of candidate picks—we can use the NOC list as a source of famous names—and [S] names a government agency, again chosen randomly from a list that contains such staples as State, Treasury, Agriculture, Health and Commerce. Since any resonance between the fillers chosen for [X] and for [S] is going to be entirely accidental, this approach will do little to inspire a reader’s confidence in the bot’s understanding of its task, and if the bot works at all as a conceit, it will be because its random choice of fillers hints at the presumed randomness of the actual political process. “Look,” it will effectively say to its followers, “the real process is as dumb as I am.” But we can imbue the template-filling process with a bit more intelligence by exploiting the relational structure of the NOC to create a quirky mini-narrative with each tweet. So what we are aiming for here is the anarchic silliness of Monty Python’s famous “Ministry of Silly Walks” sketch, in which the bot invents a government department that is very likely absurd from the get-go, but then suggests an appointment to this absurdity that seems both silly yet somehow apt. For instance, the bot can create a separate department [S] for every Typical Activity in the NOC list, from “Running a Bureaucracy” to “Providing Comic Relief,” and then fill [X] with the name of a famous person linked to that activity in the NOC, such as Adolf Eichmann for the former and Sideshow Bob or Baldrick for the latter. These pairings suggest that a president-elect is ignorant enough of government to believe that a department as silly as [S] exists, or should exist, yet is astute enough about cultural optics to pick the very best filler [X] that history, or fiction, has to offer. So our template now sheds a variable but takes on a more tightly knit internal structure as a result: “Trump taps [X] for Secretary of [X/Typical Activity].” Some instantiations of this template will, through no intention of the bot, strike more resonant notes than others, such as “Trump taps Ron Burgundy for Secretary of Maintaining Salon-Quality Hair” and “Trump taps Vladimir Putin for Secretary of Bullying Neighboring Countries,” yet even the silliest instantiations will present a scenario that is internally coherent and incongruously appropriate.
A diversity of templates can suggest different within-tweet scenarios that will be unpacked by readers to suggest varying cross-tweet narratives about a bot’s real target. To satirize an incoming administration, it serves our goal for the bot to concoct scenarios that demonstrate a certain degree of informed fantasy on the part of the president-elect and his or her team. To use the phrase coined by Aristotle, it takes “educated insolence” to construct a fantasy informed by fact and conventional wisdom, even if a fantasy is crafted to showcase the presumed ignorance and stupidity of those whose worldview it is designed to satirize. Our templates can exploit knowledge of Group Affiliation in the NOC as follows (and, if space allows, the hashtag #DrainTheSwamp can be appended to the end of each):
Trump—who mocked [X/Group Affiliation] during the campaign—taps [X] for Secretary of [X/Typical Activity].
Trump, who once promised to shut down [X/Group Affiliation], picks [X] to lead Dept. of [X/Typical Activity].
Trump, who received millions from [X/Group Affiliation]’s PAC, picks [X] to lead Dept. of [X/Typical Activity].
Trump taps [X] for Dept. of [X/Typical Activity], despite FBI reports that [X/Group Affiliation] meddled in election.
Though [X/Group Affiliation] ran a Clinton SuperPAC, Trump taps [X] to head Dept. of [X/Typical Activity].
CIA says Russians have secret tape of Trump [X/Typical Activity] with [X]. Senate calls on [X/Spouse] to testify.
This might count as “fake news” if the scenarios painted in these tweets were not so inherently ridiculous, showing more kinship to the content of The Onion or The Daily Show than to that of The Drudge Report or Breitbart News. This may be storytelling tailored to a specific target and domain, but our bot’s willingness to cross boundaries of fiction and history marks out its tales as informed nonsense. This template suggests an unhealthy mingling of reality with unreal “reality TV”:
Wanting [X:fictional] for Sec. of [X/Typical Activity], Trump is told [X/pronoun] doesn’t exist, picks [X/Portrayed By] instead.
This template turns a laudable intention into a risibly ineffective piece of theater:
Seeking to heal a divided nation, Trump nominates [X] and [X/Opponent] to jointly head Dept. of [X/Typical Activity].
The first can be instantiated as, “Wanting Jack Bauer for Sec. of Chasing Terrorists, Trump is told he doesn’t exist, picks Keifer Sutherland instead,” and the second as, “Seeking to heal a divided nation, Trump nominates Batman and The Joker to jointly head Dept. of Preventing Crime.” We want our bot’s stories to show insight into the foibles of their main character, a newly elected president, but they must also exaggerate the president’s personality to signal their own counterfactuality. The following squeezes satire from a familiar narrative of one-upmanship:
Trump passes on [X] for Secretary of [X/Typical Activity], claiming to be the real brains behind [X/Creation].
This might be instantiated as “Trump passes on Al Gore for Secretary of Lecturing about Climate Change, claiming to be the real brains behind the Internet.” Yet when a president’s pick for head of the Environmental Protection Agency openly questions man-made climate change, there is a sense that no counterfactual could ever match the real thing for counterintuitive caprice. Reality takes on the hue of “you couldn’t make this stuff up” when facts are paired so antagonistically as to create not harmony but friction. This is not zombie-like ignorance of the facts, but an impish disregard for facts so obvious the president must know them. Satire comes not from an ignorance of the facts but from a knowing disrespect for facts that are known to all or, in Aristotle’s words, from educated insolence. Our insolent bots can satirically disrespect facts as willfully as any politician, and by dialing down the unhinged whimsy a little, we can magnify the satirical effect by channeling it via the lens of a few apropos facts. If education is knowledge and knowledge serves to constrain how a bot fills its templates, we can achieve a degree of educated insolence by building even more constraints into our bot’s templates. These added constraints should steer the bot toward more intelligent picks that suggest an understanding of its task (e.g., secretaries of commerce should be experienced businesspeople, treasury secretaries are often rich) but they should not overconstrain it. They may constrain the choice of fillers to people in a specific taxonomic category or to those with specific Positive or Negative Talking Points, but because this is metasatire in action, we can still be surprised by how a bot chooses to fill our templates and satisfy our constraints:
Trump taps [X = Businessman] to be Secretary of Commerce, promises to make [X/Typical Activities] a priority.
Trump wants [X = wealthy] for Treasury Secretary, will make [X/Typical Activities] a first-term priority.
Trump promises to release his tax returns when [X = wealthy]—who made a bundle [X/Typical Activities]—does the same.
Trump taps [X = media-savvy] for White House Communications Chief, experience of [X/Typical Activities] considered valuable.
Trump appoints [X = Criminal] to head up the DOJ, brushes off a storied past of [X/Typical Activities].
Trump gifts Dept. of Energy to the energetic [X = energetic] as Sec. of [X/Typical Activities] already filled.
[X = drug-addled] to be Trump’s pick for Surgeon General; dealer connections considered a plus.
Fighting fire with fire, Trump appoints [X = ruthless] as counterterrorism advisor, experience [X/Typical Activities] a plus.
To promote the American dream, Trump appoints [X = inspiring] to be Secretary of State, looks forward to [X/Typical Activities] together.
If these constraints provide the “educated” side of the bargain, their obvious inadequacy as a filter for whimsy and absurdity provides the “insolence.” Though it is perfectly reasonable for a president to pick a business leader for the position of secretary of commerce, and to make that person’s business goals his own, our concrete choices often mock our generic aspirations, as in these instances: “Trump taps Ebenezer Scrooge to be Secretary of Commerce, promises to make pinching pennies a priority” and, “Trump wants Lex Luthor for Treasury Secretary, will make promoting greed a first-term priority.” Imagine if every mindless use of formulaic language could be exploded from within like this! Well, when a disloyal friend whines, “We were like brothers once,” you can always retort, “Yes, Cain and Abel,” or, “I know, Michael and Fredo Corleone.” If an employee seeks a raise with the dubious claim that “I do the work of two people for this company,” you can always reply, with analogical righteousness, “Yes, Laurel and Hardy.” Or if an angry spouse points to a mob of dung-flinging apes on the TV and mutters, “Your relatives, no doubt,” you might hope for the words, “Yes, my in-laws,” to trip off your tongue. Formulaic templates assume equally formulaic fillers, such as loving brothers, hard workers, and blood relatives, not the category outliers of brothers who kill or betray one other, inept workers who wreak havoc all about them, or relatives linked only by marriage. The devil of the satire lies in the disruptive details that concrete fillers bring to any blandly generic arrangement of ideas.
Yet this is far from an uncommon mode of conversational humor, for when jaded by the unthinking use of clichés by others—that is, when other people behave like linguistic zombies—we often choose to agree with an interlocutor by giving their words the most hyperliteral and unhelpful readings possible. When someone claims, “This is a marriage of X and Y,” the speaker typically assumes the marriage to be a good one and tacitly hopes that listeners will too, but we are free to respond, “Yes, a sham marriage,” “a shotgun marriage,” or even, “I give it six months.” Humor researchers actually give this adversarial form of humor the name “Trumping,” in the card-playing sense and not the presidential sense.19 Trumping is a highly productive strategy for humorous one-upmanship in cliché-laden conversations. For when one speaker lazily views language as a ragbag of prefabricated forms—clichés, idioms, templates, and what have you—the other has the rather gratifying opportunity to give new life to these hoary old forms by finding the most colorful and least clichéd instances of their lazily used ideas. When our bot relies on template filling to generate its tweets, it can rise above the level of a mindless zombie by setting out to satirically trump its own templates. This is made a good deal easier by the use of the NOC list, which, though full of familiar faces, is also home to the most colorful and extreme instances of humankind: almost any well-crafted template can be “Trumped” by filling its empty slots at the NOC well.
But what then? A template is filled, somewhat intelligently, and then our bot moves on, or shambles on, having barely scratched the surface of its own conceit. To show that our bot has the capacity to dig beneath its veneer of satire and do something interesting with its newly confected ”what if,” we might ask it to weave a whole narrative around its initial premise. So, what if Donald Trump appointed Ebenezer Scrooge to the Commerce Department, or Lex Luthor to the Treasury? Does the bot understand enough of its own conceit, and enough about the world, to tell us what happens next? As William Wallace Cook writes in his introduction to Plotto, “strike the flint of Obstacle with the steel of Purpose and sparks of situation begin to fly.” Our Twitterbots make sparks fly by doing just this, using controlled randomness to strike purposes against obstacles and protagonists against antagonists, but can they also fan these tiny friction sparks into full-fledged fires? As Cook set out to show with Plotto (in 1928!), this is easier to do than it might initially seem. Previously we presented a large collection of plot triples that, following in Plotto’s footsteps, are designed to drive a narrative forward from an initial starting action. So it is a simple matter to associate a plot action with any or all of our earlier templates, so that an instantiated template provides the very first action in a story, or perhaps one of the three actions in the very first plot triple of a story. For instance, the actions “nominate,” “favor,” “promote,” and “pick” are each salient to a template in which Trump proposes a certain NOC character for a specific government post, while the plot actions bow_down_to, curry_favor_with, impress, and pander_to seem just as apt for the filler of this post.
When our Twitterbots are built in a knowledge-based fashion from reusable resources, they become reusable resources themselves, to be clicked together like LEGO bricks when the need or the opportunity arises to build ever more sophisticated makers of meaning. By hewing to this philosophy, bot builders can nimbly react to newly topical and seasonal themes in the service of both protest and whimsy.
When reviewing Flann O’Brien’s masterwork, At Swim-Two-Birds, which espoused a thoroughly postmodern view of educated borrowing and insolent reuse, the writer Jorge Luis Borges offered this insight into mixing the real with the unreal:
Arthur Schopenhauer wrote that dreaming and wakefulness are the pages of a single book, and that to read them in order is to live, and to leaf through them at random, is to dream. Paintings within paintings and books that branch into other books help us sense this oneness.20
Had Borges lived in the age of Twitter, he might also have glimpsed a hint of this oneness in the interwoven tapestry of a Twitter time line. If Twitter as a whole is a vast book of tweets, some embedded in others and some branching endlessly into viral retweets and impassioned replies, we each get to sample just a small subset of its pages, no matter how many accounts we may follow. The notions of order and randomness are interwoven too, for though we cannot predict which accounts might tweet next, Twitter forces them all to march into our time lines with the orderliness of an English vicar queuing for stamps. In this way, the mundane and the whimsical are audaciously mixed up, as the outpourings of celebrities, politicians, comedians, news services, friends, coworkers, and bots march to the same Twitter drumbeat. So truth rubs elbows with half-truths, lies, and Frankfurt-style “bullshit,” while dreams and fantasies mingle with hard reality.
Bots such as @pentametron contrive to reorder the pages of the Twitter book so that mundane wakefulness becomes poetic dreaming, by nudging readers to perceive a semantic or pragmatic resonance between tweets that are paired on metrical grounds only. It works much like a dating agency that pairs members on the principle that it is the couples with similar height, weight, or other superficial measures of compatibility that generate the most electrifying sparks. In this way, the bot adds value to tweets that may have already run their authors’ intended courses. Indeed, by striking sparks from what might otherwise seem like spent fuel, our bots confront head-on the specter of disposability that haunts not just Twitter but older forms of content delivery too. The cultural critic Mark Lawson, who writes extensively about television, diagnosed TV’s pre-HBO lack of artistic standing with an insight that seems as relevant to Twitter now as to TV in the 1980s and 1990s: “The invisibility to posterity has always been television’s difficulty. Many programs are intended to be disposable, to disintegrate even as you look at them.”21 Even the wittiest tweets are disposable ephemera, flashes of light that quickly recede as our time lines fill with new content. Our bots only quicken the pace with which they recede from view and from memory by using automation to ensure that new content is produced with clockwork regularity to supplant the old. Our bots can no more hold back time than King Canute could hold back the tides, but that has never been their purpose. Bots can give new life even to content that is intended to be disposable and take from their disintegration the material for new tweets. Moreover, as our Twitter lathes produce their steady streams of sparks in the form of whimsical what-ifs, these may in turn ignite the imaginations of human users (and perhaps other bots) who might then refine, repackage, or satirize these ideas in a never-ending cycle of disposable creativity. Individual tweets may be disposable, but the overall cycle of creativity lives on.
We build our Twitterbots to be tourists in strange lands. We set them loose to explore those pocket universes on our behalf and to send us frequent postcards on what they see there. These are realms of pure imagination, not of hard reality, but they are worlds that often mock our own with their simplicity, freedom, and elevation of form over meaning. Fans of TV’s Rick and Morty might see parallels here with Rick’s interdimensional cable box that allows him and the Smith family to watch an infinitude of inventively awful (yet bizarrely attention-holding) shows from across the multiverse. Here is the show’s introduction to an oddly familiar Saturday Night Live! that is wildly popular on another world:
It’s Saturday Night Live! Starring a piece of toast, two guys with handlebar mustaches, a man painted silver who makes robot noises, Garmanarar, three s-eh-bl-um-uh-uh-uh- I’ll get back to that one, a hole in the wall where the men can see it all, and returning for his twenty-fifth consecutive year, Bobby Moynihan!22
This may sound like TV made by Twitterbots, but who wouldn’t want to channel-surf shows like these? Other briefly glimpsed shows from skew-whiff universes of S1E8 include a Games of Thrones, where everyone is a dwarf, except, of course, for the vertically challenged Tyrion Lannister, and a poorly paced true crime show, Quick Mysteries, that reveals all of its cards up front. The rapid-fire invention of Rick and Morty reminds us that we humans are the universal what-if machine, the mental equivalent of Rick’s interdimensional cable box, capable of inventing endless new worlds to visit. Though our Twitterbots may be far from universal, we can build them to explore bespoke new worlds on our behalf, to dig deeper than our schematic view of the world and its rules might otherwise allow, to ferret out the weirdest instantiations of these rules for us to sample and enjoy. So while our individual bots may resemble a Bizarro channel on multiverse TV, with its weird tics, whacky obsessions, and view-askew take on life—for instance, despite the seriousness of @NRA_Tally, its four-hourly killing sprees can read like a deliberate parody of modern cable programming—collectively they turn Twitter into Rick’s interdimensional cable box, allowing us all to channel-surf the wonders of a multiverse where humans are just one voice among many.
The various resources described in this book have been designed to allow bot builders to respond nimbly to new what-if opportunities for Twitterbots as the holidays, the seasons, and changing circumstances present them. They allow our Twitterbots to bring a quirky understanding of this world to their automated explorations of other worlds, to help them appreciate what they find there, and to help them to better filter the noisome chaff from the tweet-worthy wheat. By using knowledge, scant though it may be, to lend some familiarity to the oddities of an artificial world, they also inevitably show us the strangeness of our own. This is what Twitterbots do best: they remind us of the strangeness of language, social convention, and human nature more generally by allowing us to see familiar human qualities in the synthetic, the mechanical, and the alien. In your onward explorations of this multiverse of possibilities, do consider docking occasionally at BestOfBotWorlds.com, to share your experiences with others and to refuel on resources and ideas. In the final analysis, it is not the Twitterbots but the Twitterbot builders that make Twitter the best of bot worlds.
In the spirit of opening new doors while closing others, we conclude this final chapter by building the ultimate what-if machine: an interdimensional cable box of our own. We can glimpse the possibility space of television in the tweets of a restless channel-hopping Twitterbot, which we will build by repurposing a variety of generative components from the Tracery grammars of other bots in earlier chapters. As good a place as any to start the construction of our bot is a mainstay of the TV viewing experience in all dimensions: advertising.
One of our earliest Tracery grammars from chapter 3 exploited the power of raw combinatorial generation to coin new words and new meanings from the collision of Greek and Latin roots and their standard interpretations. As the classical roots of these new words are often suggestive of the kinds of products we might discreetly seek out in a pharmacy, our neologism grammar is easily converted into a generator of faux-scientific gizmos and doodads. Our product pitches will make for more compelling TV if we recruit the famous faces of the NOC list to act as celebrity shills, as in this commercial reframing:
Ernst Stavro Blofeld swears by “GaleoMart.” When you need a place of business dedicated to the sale of sharks, there's none better!
This pairing of sharks with a Bond villain is merely an accident of random generation, yet we build our bots to foster such happy accidents. To cultivate many more combinatorial delights, we can repurpose our Just Desserts grammar from chapter 3. While those treats were created to be vengefully vile, perhaps this is how the denizens of other worlds actually prefer their desserts. By also defining a set of apt prefixes and suffixes to combine, we can generate names for the companies that make the awful treats, as in this grammar output:
Why not try TrumpDessertz-brand Peanut butter cups made with used coffee grounds instead of chocolate—Feed your desires!
Famous people can also provide celebrity endorsements for the services of a company with deep pockets, and a diverse source of possible services is to be found in the moral maze of action frames and roles we explored in chapter 4. We can repurpose our moral grammar to generate advertising such as this:
When Lord Voldemort wants to commit top-notch killing he calls 555-Predators. They won’t be beaten on price
In chapter 7 we built a grammar to map the normative properties of familiar visual ideas to their related dimensions. This allowed us, for example, to pitch the dull gray of solid rock as the representative color of solidity itself. The same approach can be used to lend a familiar face to more abstract dimensions, as when we reinvent a NOC talking point as a celebrity perfume:
Parfum de “Reclusiveness”—the new scent for men from JD Salinger.
These four generators combine to form a single Tracery grammar named Advertising grammar.txt in the Interdimensional Cable directory of the TraceElements repository. This grammar provides the foundations for the rest of our dimension-hopping TV service, on which we will now build another fixture of the cable landscape: the twenty-four-hour news cycle. Any news service needs newsworthy propositions to broadcast, and even assertions about the most famous people need interesting claims at their core to reach the air. Darius Kazemi’s @twoheadlines is ideal in this respect, as it is constantly fueled by up-to-the-minute headlines on the web. However, to squeeze our news generator into a Tracery bot, we shall have to confine ourselves to a closed-world model of newsworthiness. Fortunately, as we saw in chapter 9, dbpedia is a rich source of conversation-worthy categorizations that can be as informative as good gossip and headlines. By harvesting every dbpedia category that matches any of the patterns X_who_have_Y, X_who_were_Y, X_who_can_Y and X_about_Y we can extract the central claims (the Ys) around which our bot’s news headlines will be based. It then remains for our bot to invent a suitable name for the cable news network in question. Using Fox News as our inspiring exemplar, we leverage our list of animals from chapter 3 to create other animalistic news stations for the multiverse, as in:
JaguarNews Exclusive: Vladimir Putin has nothing nice to say about The Irish Mob. Up Next …
MonkeyNews Exclusive: Martha Stewart denies claims she has acquired Austrian citizenship. Stay tuned
News stations are also prolific producers of documentary films, for which the dbpedia categories matching X_about_Y provide some topical Y themes:
Coming up on HogNews, tonight's documentary, “Coups d'État,” narrated by your host Barack Obama
What about the multiverse of fictional content to be found on interdimensional cable TV? The selection of films and scripted shows will be infinitely diverse, yet as weird as any might be—and unlike most of the books in Borges’s Library of Babel—the premise of every film or show should present a veneer of sense. We can easily create a scrambled version of our own TV reality by harvesting a large list of movies and shows from dbpedia to randomly recast them with random selections from the NOC list. But a more interesting source of weird and wonderful titles are the creative (and often ironic) similes we harvested in chapter 5 using the pattern “about as [X] as [Y].” The Ys we obtain from these similes are typically chosen to be both visual and ridiculous, making them well suited to reinvention as wacky cinematic vehicles with apt (or deliberately inept) casting choices. Consider the simile “as musical as a box of smashed crabs.” Repackaged as an ill-advised movie, this becomes:
Stay tuned for “The Box of Smashed Crabs” starring Eddie Van Halen and an especially musical squid. Up next …
The exaggerated qualities of inventive similes lend themselves nicely to a satire of Hollywood’s most outlandish offerings, as in this grammar output:
Stay tuned for part 2 of “The Hot Tub in a Limousine” starring Bruce Wayne and an especially decadent locust. Up next …
At the same time, similes allow the NOC-based grammar to respect the logic of its own bizarre choices, choosing Eddie Van Halen for a musical vehicle and Bruce Wayne for a film dripping with decadence. You can find the entire grammar (advertising and news and fiction) in a file named Interdimensional grammar.txt that, although large, is amenable to manual editing. How might you rewire its parts, reframe its outputs, or add to the scope of its ramblings? Will you add a subgrammar for infomercials, perhaps, and another for satirizing the excesses of reality TV? Or a subgrammar for generating alien names (such as “Garmanarar”) for the celebrities and actors of the multiverse? Perhaps your new additions will be sufficiently generative to spawn a whole new bot of their own? The world of bot building is as broad and accommodating of new conceits as any interdimensional cable box, and one where our wildest metaphors can easily become playful reality.