“@KarlMarxovChain” and “@MarkovChainMe” by Moacir P. de Sá Pereira

These two bots use Markov chain generators to produce tweets, but distinguish themselves from other bots by responding to input from those who interact with them via Twitter. (Click on the images below to visit their pages).

Screen capture from "@MarkovChainMe" by Moacir P. de Sá Pereira. Twitter profile of @MarkovChainMe displaying his two most recent tweets. Text: "If you @ me, I'll @ you a chain built from your most recent (100) tweets. If you "@markovchainme do @user", I'll use @user's tweets. @muziejesus wrote me. Probability Field / @ryansroberts sailing past on at work to me with names like "krusty" you'd have been getting everyone in real fucking waste of course / @rhmgroo care, rather actually provide care, rather than ignore care It is just a point. lie. and wish louise mensch would learn"
Open “@MarkovChainMe” by Moacir P. de Sá Pereira

“Markov Chain Me” is like a curved mirror that shows you a generated version of your Twitter identity. This is a fascinating way to analyze your own writing, getting a sense of what you’re likely to tweet about. Even better, you can direct it to analyze someone else’s tweets by using the phrase “do @——.” Hopefully my request that it “do” itself doesn’t cause it (or Twitter) to somehow implode. The response time is fairly quick and recognizable: one can sense the original voice in that tweet, which is part of what is so fascinating about the method.

Screen capture of "KarlMarxovChain" by Moacir P. de Sá Pereira. Twitter profile of @KarlMarxovChain displaying his two most recent tweets. Text: "If you tweet me a word, I'll use it as a seed and respond, either from early works or Das Kapital. Otherwise I seed myself 5/day. @muziejus wrote me. / Ages, did not rise above the system of land to public purposes. A heavy progressive or graduated income tax. / England and Naples after the fall of State and the petty bourgeois has been launched, the proletariat would derive from a fetter"
Open “KarlMarxovChain” by Moacir P. de Sá Pereira

The Karl Marxov Chain responds to a word that users (or Pereira) seed it to guide its search through Karl Marx’s publications, as described.

When it gets the seed word, it finds it in the text and takes not the next word, but the next two words. The first two words of this 3-gram are first two words of the tweet. It then takes not the last of these words, but the last two and searches the text for that pair of words. Then, of all of the times that those words appear together, it picks one at random, adds the last word to the chain, and then moves up a word. The result is that the probabilities are a bit more constricted, meaning that the tweet conforms a bit more closely to the original text, meaning it ends up sounding a bit more like normal English.

The bot also cheats a bit and tries to make “complete” sentences (start with a word that has an initial capital in the source text and end with a period), but it’s not always successful. The source texts are also not the cleanest in the world, so it sometimes hiccups and tosses out typographical gibberish.

In practice, it seems to create semi-random mashups of different sentences by Marx, giving it more of a cut up feel than other more granular Markov bots, like “10 Print ebooks.” Its pun-induced name is well within the tradition of other mashup bots, like “Justin Buber,” “Kantye West,” and “Kim Kierkegaardashian.”

It’s about time we used digital tools to deconstruct our philosophers.

Featured in Genre: Bot