The art of the Twitter bot

There are a lot of voices on Twitter and not all of them are human. In a quarterly filing with regulators at the beginning of November, Twitter disclosed that 8.5% of its users showed signs of being bots. With 307 million users in the third quarter, that makes 26 million accounts more robot than human. That’s a lot of bots talking side-by-side with real people.

George Buckenham (@v21) is a game developer, bot wrangler and the man behind Cheap Bots, Done Quick! – a site that helps users make their own Twitter bots. I meet Buckenham in the courtyard of Somerset House, where he is working on a physical-digital-animal-stacking game called Fabulous Beasts.

There’s a small crowd of ice skaters circling a rink close to where we sit, and Buckenham talks to me about internet noise.

“The thing I like is the overloading – the semantic satiation that you can get with words. There’s this constant feeling you get when you’ve read too much internet, where it’s an overload of facts and opinions and blah and Hot Takes coming at you. And then, suddenly, you have a bunch of robots creating fake ones as well, at the point when you’re exhausted and overloaded with that stuff, suddenly there’s even more. It’s the aesthetics of excess – of too many opinions.”

@thinkpiecebot is a good example of a bot that adds an automated voice to the daily influx of opinions, using a series of randomly selected formulas and predetermined lists of words. This results in headlines like “Reverse Racism: The Real Issue Facing Syndicated Columnists” and “What Does George Lucas Mean For Craft Beer?”

While these bots satirise the culture that’s built up around Twitter, Buckenham tells me about the beauty of bots that stick out from the echo chamber. “You read through Twitter and it’s this constant mass of hashtag content, and it’s really nice to mix in an artwork. Instead of feeds of people sharing stuff it’s like: ‘Here’s a moth. Here’s a beautiful moth that a machine has generated.’”

Accounts such as @mothgenerator, @GenerateACat and @NiteAlps all make procedurally generated artworks, while @MagicRealismBot performs a similar trick for magical realist stories. Follow these accounts and you’ll soon find fragments of machine-made art spliced between human sentences. But why are these automated acts of imagination so appealing? Perhaps there’s something subversive about these bots, which digs a handful of fingernails under the fabric of Twitter and lifts the lid on our intermittent expressions.

Then there are bots such as @DeathMedieval, @everybird_ and @everyword, which present an regular feed of ordered information – real deaths from medieval coroner reports, bird names and every word in the English language respectively.

Buckenham tells me there is a conceptual appreciation to be had knowing that these accounts exist – that “the joke of it is as important as the actual existence of it” – but it seems to me there’s also something to be said for how words and historical deaths can be appropriated and recontextualised by retweeting, quoting and responding to these tenacious indexers.

Buckenham tells me that another aspect of Twitter bots is the enjoyment of play-acting that these bots are people. He tells me about the #botALLY hashtag, which draws members of the bot community under the playful banner of friends-of-bots.

“I don’t think it’s as successfully impersonating a human, it’s more just existing on an equal footing as a human – that the tweets are no less important than a human’s tweets,” Buckenham tells me. “It’s not a second class thing. The idea that Twitter is a commons that you can share with these alien, non-human entities that have their own rules. Where you treat them seriously as humans and play acting that they have rights and responsibilities. It’s all tongue in cheek, but at the same time it’s more interesting to engage with things that way than dismissively I think.”

One project, NaNoGenMo, takes this playful approach and runs with it. Instead of NaNoWriMo (National Novel Writing Month), which aims to help people write a 50,000 word novel in one month, NaNoGenMo (National Novel Generation Month) encourages users to spend the whole of November writing code that generates a 50,000 word novel.

The result is often a Borgesian cacophony of possible words and characters. One resulting project, Alphabetical Order by Leonard Richardson, was generated by searching 47,000 plain-text public domain works for lines that contained no alphanumeric characters.

(Above: An extract from Alphabetical Order by Leonard Richardson)

Even if bot-made artworks may ultimately be non-human, the responses people have to their tweets can be very human. @youarecarrying is an account that, if you interact with it, gives you a series of inventory items. It then encourages you to draw the objects and post them alongside the hashtag, #iamcarrying.

Like retweeting lists of words from the dictionary, the fun of interacting with bots comes from humanising the machine. For bot makers, there’s joy to be had in throwing a self-perpetuating voice into the online cacophony, then seeing that voice scooped out of its original procedural context and re-framed elsewhere by real, flesh-and-blood people.

“A lot of bots feel like they are artworks that aesthetically respond to the way of reading on the internet. Especially ones that are purely creating noise and static, and comprehensible phrases mixed in with unicode garbage. I made a thing that creates animated gifs of random noise, and there’s this wonderful thing where I go on Tumblr and I scroll through and see art that’s been created on my server every hour. There’s a magical-ness to that, that a computer’s doing it on its own.”

Buckenham tells me that, for bot makers, there are ethical responsibilities for what bots produce: “A flipside of #botALLY – pretending that bots are people or treating them with equal respect as you would people – is the idea of taking responsibility for the things your bot posts,” he says. “There’s super ethical stuff that comes up, which is difficult to negotiate if you’re generating text. If you’re generating random text by, say, ripping stuff from headlines, then you could very easily end up with some horrific stuff. It’s your responsibility as a bot creator to not tweet hurtful things.”

I ask Buckenham whether there should be an editorial process to bot output, or whether there is something to be said for letting bots carry on without human control. “I think there’s various ways to approach it, and lots of them are valid in different ways,” he tells me. “But I think the main overriding principle is you’ve got to take responsibility. It’s your bot, you have control over it, so you’ve got to take responsibility for it. It’s as if you tweeted it, because in some sense you did tweet this.”

By way of example, Buckenham tells me about a number of accounts that walk close to the line when it comes to ethical dubiousness. One of these accounts, @NeedADebitCard, retweets these pictures people have posted of their credit cards. It comes with the tagline “Please quit posting pictures of your debit cards, people,” and at the time of writing has 17.7k followers.

He also tells me about @FFD8FFDB, an account that posts images from unsecured camera networks. “It’s a total minefield in that it’s taking stuff that people aren’t even necessarily aware they’re posting to the internet in terms of a webcam feed, and then posting it up to Twitter.”

Derek Arnold, who is behind the account, has written a treatise on the project. In it, he explains that he excludes cameras in private spaces, and that the focus of the project is on the aesthetics of the images rather than their intrusiveness.

“I enjoy that strangers find it unsettling, amusing, or even uninteresting,” he writes. “Like other Twitter bots, its unending tenacity is part of its charm. Many cameras go dark at night, most not having enough illumination to provide images. The bot doesn’t care and keeps stealing shots.”

(Above: A shot from Derek Arnold’s @FFD8FFDB account)

The shots posted on @FFD8FFDB do indeed have an eerie sense of beauty, but the voyeurism inherent in their capture and depiction is nevertheless largely responsible for this. It’s a million miles from the progressive listing of something like @everyword, and yet there’s still a sense of alien detachment to the regularity of the process – a feeling that beneath the human retweets and replies is a machine habitually clicking through teeth on a gear.  

Invasion of the body snatchers

With bots spanning everything from fake magical realist novels to insecure webcam feeds, I ask Buckenham how he thinks the bot scene will evolve over the next few years.

“My personal hope is that it splinters and people go off and do their own stuff, that it doesn’t just go in one direction, but goes in different directions and deepens out,” he tells me. “The range of people making bots is broadening and that’s super interesting in terms of the different ways people use them.”

Once the technical barriers to bot making have been lowered – with the help of sites such as Buckenham’s Cheap Bots, Done Quick – it will indeed be interesting to see what bots are made beyond those belonging to code-savvy artists. This could give rise to a gamut of interesting projects, but it could also lead to an inundation of automated fictions – an Invasion of the Body Snatchers scenario where coded simulacrums outnumber real people.

Social networking apocalypses aside, for now I’m more than happy with procedurally generated moths and historical deaths butting in on my Twitter feed. With all the voices clamouring for attention every time I use my phone, some of the words I enjoy most don’t come from human mouths.

Next: What’s the point of virtual reality?