Social Media Narrative:
The Rutgers Camden Digital Studies Center
Facebook, November 16 - 21, 2016
C hris Rodley is a PhD candidate at the University of Sydney, where he is studying the impact of social media on writing. He also teaches digital media classes there and at the University of New South Wales. Chris's work at the intersection of literature and technology includes the flash fiction generator Magic Realism Bot, co-authored with Ali Rodley; and the data art projects Everything Is Going To Be OK, B.E.T.T.Y., and Death of an Alchemist, which were co-authored with Andrew Burrell. He is currently a contributor at BuzzFeed.
The Magic Realism Bot
If you're on Twitter a heck of a lot, you might possibly have seen my bot Magic Realism Bot, which is a collaboration with my sister Ali. Every two hours, it tweets a randomly generated micro-story in the style of writers such as Jorge Luis Borges, Gabriel García Márquez and Isabel Allende. I wrote the syntax, phrases and vocab lists, but not the code. Here's the Facebook version, though it's a pale shadow of the original: Magic Realism Bot
(People sometimes ask how it works. It uses a template-based technique akin to Apples to Apples or Cards Against Humanity: I wrote general, abstract syntaxes with all the key words missing and the code fills in the blanks by selecting from very carefully tagged vocabulary lists. But there's lots of wild cards and other features to make the results more complex, coherent and variable.)
I could talk about lots of things but here's four related issues that I often think about:
1. Every day there's lots of warm, convivial interactions in the mentions of the bot, and I don't think I can ever remember seeing any trolls or aggression. Check out this Twitter search of the bot's mentions to see how users are responding to it in real time:
2. A common reaction is for people to relate the bot to people and events in the news, which meant it took on an odd role in refracting the US election season through a magical, otherworldly lens. Sometimes the connection was fairly obvious -- it's easy to see Hillary Clinton in "A blonde lawyer falls in love with late capitalism". But people also related apparently abstract tweets to politics, such as "A clairvoyant turns over a tarot card with a pianist on it. 'Your destiny is to become a gas station attendant,' she says to you" which people interpreted as a comment upon widening inequality.
3. Another common reaction is for a user to relate a tweet to their own lives, or to someone in their social circle. I like to think of this as "botomancy", a digital form of bibliomancy (the old-fashioned method of fortune telling by consulting a random line in the Bible). Botomancy is a way of finding out about ourselves using the random recombinations of bots. I've noticed this phenomenon going right back to Allison Parrish's @everyword, the Australopithecus of the Twitter bot's evolutionary history: what seemed to be simply an alphabetical list of words was quickly embraced by users as a way of understanding themselves and their lives (LGBT users later embraced the spin-off bot @everywordisgay in the same way). I'm not quite sure what to make of this phenomenon, but I wonder if it points to the value of randomness in an increasingly rationalized, algorithmic world? It's also a reminder that in social media e-lit, understanding the context and audience reception is even more important than in traditional lit. A critic who only saw the tweets of @everyword would have zero understanding of the incredible responses the work catalysed.
4. As I've shown, the human audience is a huge part of machine-generated literature. Another big part, of course, is the human authors! One of the most striking things about the early responses to Magic Realism Bot is that people would seemingly assume the stories arose from scratch, from a tabula rasa, perhaps via a deep learning engine which had learnt the DNA of magic realist stories by munching through thousands of them. I'm sure that will appear sometime in the 21st century, but machine-learning recreations of narratives are preeeety crappy compared to the human version. One of the key aims in Magic Realism Bot was to make a bot which relied not on technically innovative processes per se but simply aimed for innovative results by any means necessary (except for cheating by writing, editing or curating tweets, which I'm sometimes accused of but am far too lazy for). This was achieved through clever coding but also some workaday, traditional-ish human authorship: I had to come up with lots and lots and lots of different magic realist templates and small variations in those templates. Other botmakers are often surprised at how big the database is and how often we update it, because the culture is very much to create "set and forget" bots (one exception is Nora Reed who regularly updates the wonderful @thinkpiecebot).
I say all this to make a much more general point: we don't need to think of these bots in a dualistic way as either machine (eg @oliviataters) or person (eg @horse_ebooks). Rather, bots on social platforms are hybrids of authors, their tools, and the responses of other social media users (usually the readers, but sometimes also other random users whose content has been swept up and spat out again). I wonder what other kinds of literary machines we could create if freed from the obligation to regard them as fully automated entities?
Transcript of The Magic Realism Bot Conversation