About a bot: Interview with Katie Rose Pipkin


@mothgenerator by Katie Rose Pipkin and Loren Schmidt

Taina Bucher interviews artist and bot maker Katie Rose Pipkin about her most popular Twitter bots, how they work and what they mean. Indeed, what are bots, who else is engaged in artistic bot making, and how will social media bots evolve?

Meet Tiny Star Fields. Several times a day, the Twitter account publishes a field of stars in different shapes to a dedicated 51.000 followers. The latest tweet, published 53 minutes ago, has already been retweeted 151 times and gathered 114 favorites. Tiny Star Fields is a Twitter bot. During the last few years bots, or automated pieces of software, have becomes an integral part of the Twitter platform. As some recent reports suggest bots now generate as much as twenty-four per cent of posts on Twitter, yet we still know very little about who these bots are, what they do, or how we should attend to these bots. Admittedly, star-tweeting bots like Tiny do not belong to the kinds of bots that are most talked about. When people usually think of bots they mostly have a specific type of bot in mind, one that animates feelings of annoyance and disturbance. The spam bot, however, is but one kind of bot.

As Tiny, and many others like her attest to, bots are just like people. They are different, tweet for different reasons, have specific audiences and engage with the world in various ways. Guided by their human programmers or taught to learn from existing data in playful ways, bots are legitimate users of platforms. But bots would be nothing without their creators, their makers who have conceptualized and brought these digital personas ‘to life’. So let’s not just meet Tiny Star Fields, but also Katie Rose Pipkin, the 24-year old artist and creator of Tiny Star Fields.

Katie, why don’t you tell us briefly about yourself and your background?

I grew up in the woods of Austin, Texas, where I also attended university for my undergraduate degree in studio art. Most of my work there was focused on drawing and installation, but I was also curating internet ephemera and beginning some rudimentary code projects at the time (albeit in isolation from others doing similar work). I also have a history in curation, and have run creative spaces for many years. I’m currently pursuing my MFA at Carnegie Mellon, in Pittsburgh.

What got you started with making Twitter bots?

I started making bots in the summer of 2014. I had moved to a very tiny town in rural Minnesota (population 900) for a longer-term artist residency, and was quite isolated. I didn’t have a car, there was no bus or train, and I didn’t know anyone there. I was used to being alone on residency, but often I had friends near enough to visit or a local coffee shop to haunt. Here, with no other options, I was home and online almost constantly. The internet has always been important to my practice (and my social life), but I really attached myself to it as a lifeline in that period.

I was already following twitter bots (@everycolorbot, @youarecarrying, @twoheadlines, @minecraftsigns, @oliviataters, @prince_stolas and I’m sure many others), but being online constantly shifted how I thought of them; rather than just seeing their occasional statements as charming non-sequiturs in a human space, I started to notice their underlying personalities, the structure of code that differentiated one from another; when they posted, the kind of source materials used, how they interacted with others. With nobody to keep up with locally, I also began sleeping in erratic structures- some nights for 5 hours, others for 14. As a side effect, I would catch off times on twitter, where everyone but the bots were asleep. These timelines of automation had a striking effect. I was particularly fond of the bot chorus around the turn of the hour- bot ‘o clock, as some call it.

I had been following and aware of @negatendo’s #BOTALLY posts (a sort of # organizing structure for bot-related news and resources) for a while, but I also started following @thricedotted, @inky, @beaugunderson, @tullyhansen, @aparrish, @boodooperson and @tinysubversions (and many others!) in this period. There were new bots almost every day, all unique, and I was really taken by the ways in which people interacted with them, and how they operated in that social space.

How did you go about making your first bot? 

I got node.js setup on my laptop (no small task for me at the time) and figured out some fundamentals of text manipulation in javascript. After a serious number of false starts, I made my first bot, @feelings.js, in an afternoon. I made @tiny_star_field 5 days later, in the middle of the night, hiding in my basement during a tornado. The power was out and I’m almost certain I got the structure done in one laptop charge. I deployed it when the power and internet came back the next day.

You waited for the sky to clear and become sprinkled with starts again. In the meantime you made your own digital sky, that’s cool. Did you do a lot of programming before starting with making bots?

I suppose that depends on what you mean by programming? I had worked in and around browser-based experiences for years, but I had never taken a structural approach to learning code. Every new idea and project had a particular set of problems that I attacked with utter naivete, writing vast messes that were honestly pretty shocking when they worked. Looking at my source code for those projects now is very much like looking at an outsider-art approach to computer science. Which is, I suppose, what they are.

I still sometimes struggle with basic concepts just because I haven’t run into them before- I learn best when directed at a goal, and sometimes those goals skirt fundamental structures. My knowledge is a funny hodge-podge assemblage of extremely difficult concepts I needed for some project or another, while I may forget the syntax for a basic sentiment. I keep telling myself I’ll read a book or take a course on putting code together properly, but so far I just keep learning what I need. I am sure I will feel the same way about my current projects in a year or two as I do about my older projects now. My first bots are very embarrassing inside and it has only been a year and change.

You’ve said that @tiny_star_field is your most popular bot, but that your personal favorite is @feelings_js. Would you care to elaborate?

Neither of those bots came from a particularly well-considered place technically; they were the first I made and I was learning. I think I was tickled by the idea of a bot that did nothing but emote; it seemed like a charming inversion of the coldness that often creeps into automata. Tiny was a simple reflection of my unicode character habit; I have a hobby of making little vignettes or dioramas with combining characters and atypical symbols and I have been enjoying automating them. (I am also now a member of the Unicode Consortium and am working with these characters structurally.)

That comment about favorites was from a long while back and my favorite bot is probably now Moth Generator (@mothgenerator), which was a collaboration with @lorenschmidt. Its different than many of my bots; it’s really just a wrapper on an image generator; but it is the first bot I’ve made that I felt used @-replies in a truly useful manner. It takes the text of the tweet sent to it to seed the generator with a unique number; therefore, the ‘moth’ moth will always look like every other ‘moth’ moth, while a ‘bot’ moth would shift in many ways. A ‘moth bot’ moth would share characteristics of both.

How do these bots work?

Feelings.js (and a few others like it) is basically a fill-in-the-blank Wordnik wrapper. It has a variety of possible sentence structures on a switch statement, and then pulls parts-of-speech from the dictionary API. I have a few structural rules in place that slightly favor alliteration and a few other cute tendencies (as well as blocking offensive words), but it is basically madlibs.

Tiny is even simpler; it has a large array of star and space options and pulls randomly from the available options. The biggest challenge was just finding an ideal balance between character frequencies. I tweak it occasionally and don’t feel that it is quite ideal yet. I am tempted to make it sparser. I am also in the process of making a Tiny Star Fields clone that uses actual astronomical data at varying scales, so the tweets will be a literal patch of sky.

Some of my other bots are a little bit more complicated- Moth Generator is a wildly long drawing routine in Javascript, Sea Change (@100yearsrising), tweets unicode characters mapped to sea-level rise predictions over the next century. Others use more obscure text manipulation techniques and large corpuses. But I think it is important to note for folks just starting that complication does not necessarily make them stronger artistically, or more popular socially- the best things are almost always just good ideas.

What has been the response to your bots?

There is certainly an audience of bot-appreciators; sometimes I will see people who follow 30 or 40 bots but none of their makers. Bots also have their own secret lives outside of intention. Tiny auto-followed people back for a while (something like the first 6k), and this made for a truly wonderful sample! Very few of them are in the bot-community; I think the vast majority are One Direction fans. It is a fascinating slice of social lives I would never think to seek out myself.

What is the bot community that you are referring to?

Gosh, what is the bot community, good question. I suppose it seems to be a loosely associated group of folks interested in social bots, for whatever reason? People seems to come from all walks- programmers, game developers, linguists, writers, artists, analysts, poets. Making the skeleton of a Twitter bot is a fairly simple exercise and doesn’t inherently have the high knowledge overhead of some other creative programming tasks. They are also incredibly flexible in content and process and I think that mutability allows a certain wealth of intent from bot to bot. These two avenues of openness mean that they are used for all sorts of things! As entities, they are as unique as the people who make them.

                       liminal algorithms (generative text, drawing machines)

In general, I’ve found folks who are organized around making bots to be nothing but supportive, kind and interested in helping others get started with producing their own work in this realm. Within that community structure are also all the folks that might not make bots (yet) but know what they are, and are interested in their processes, or write about them, or consider them valid as artworks or creative entities.

What, indeed, are bots?

What are bots? Gosh, this is an even better question than the one about bot communities. So, there are a lot of ways to think about bots, and in my opinion they are stackable and do not refute one another. But here are my thoughts:

Firstly, they aren’t new- automata has been around for a very very long time. One can look at examples of clockwork machines or candle-powered toys from over 1000 years ago. Even beyond physical examples of automata, the idea of bots is pervasive culturally; stories about golems and enchanted armor or physical objects imbued with personality have been with us since stories began.

Digital bots (especially bots that live in social spaces) fit into this long history of objects granted almost-humanness. They fill in for a part of human action, the slice of person granted to digital representations of ourselves. Just like the golem that guards a passage, their tasks are programmed, but because they do these tasks on their own (guard, tweet) we grant them entity. Perhaps this is as much doubt (“Is it /really/ a bot, though? Maybe it’s just a person pretending?”) as it is gift.

Secondly, I do think there is an aspect of doubling or mirroring that these bots employ. They are widening the reach of their creators; they are automated versions of a specific slice of their creators. Many many bots fall into this category. Something Darius Kazemi once said first got me thinking this way. It was advice to a want-to-be botmaker who didn’t have an idea for a bot. Darius suggested ‘come up with a joke that is funny but formulaic and automate it’. This type of repetitive production is not just seen in joke bots, but almost all bots that are not attempting to emulate humanness. The maker would have made the joke once; by making a bot, it is made many times (but also, perhaps made better than it would have been the once).

To expand; the goal of work-by-generation is a fundamentally similar, but shifted process from that of work-by-hand; rather than identifying and chasing the qualities of a singular desired artwork, one instead defines ranges of interesting permutations, their interpersonal interactions, how one ruleset speaks to another. Here, the cartographer draws the cliffs that contain a sea of one hundred thousand artworks. And then one searches for the most beautiful piece of coral inside of their waters.

So, I suppose this is where bots are truly interesting to me. Because this kind of making (the looking for the best moment in a sea of automated possibilities) is a methodology of construction that feels, in some ways, new.

I like the notion of bots leading secret lives. Are you ever not in control over your bots? Or what does this secrecy entail?

I take a pretty lax approach to keeping up with my bots. I almost never log into their accounts or closely monitor what they are up to. I censor certain words I find offensive, follow them on my own twitter account, and hope I catch it if they break. This means that their notifications never reach me; the things that are said to them (or their own replies) are often invisible to everyone but them.

In what ways do people or other bots interact with your bots?

Most (although not all) of my bots are non-interactive, meaning that they do not @reply back when spoken to. That being said, they are absolutely interacted with. Tiny star fields in particular gets a ton of messages; lots of people will have conversations in the mentions. I find them pretty charming and will occasionally peek at what people are saying to one another. Since I generally keep @replies off, I don’t get the bot-to-bot eternity loops that you’ll see sometimes with the image bots or ebooks bots or others that reply. But I always like it when spam bots or reddit bots find mine by keyword search. The best example designed bot interaction might be Eli Brody’s tiny astronaut (https://twitter.com/tiny_astro_naut) which inserts spaceship emoji into Tiny star fields’ tweets, or its conceptual sibling, tiny space poo (https://twitter.com/tiny_space_poo).

How many of your bot’s followers do you reckon are other bots, and is bot to bot interaction different to how humans interact with bots?

I haven’t done the numbers, but it seems like there is a slightly higher percentage of bot to bot followers than human to bot. I would guess this is a combination of auto-following routines and being manually directed to follow entire lists of other bots. Perhaps also they are more patient with repetitive or nonsensical tweets, and stick around longer.

Most bots now have conversational abandonment built in, but this was not always the case- it was once pretty common to see two replying bots get into a conversation with one another that would last hours or days, to the tune of thousands of tweets, one every few seconds. I once got accidentally caught in the mentions of one of these cycles and had to wait for one of the bot’s owners to wake up and reset their servers. It was amazing and I also had to turn off all notifications on every device I own.

Now, I think most bots use more intelligent replying- just to one person, or randomly across their followers, or only every 10 hours, or perhaps replying to keywords or requests. This has made bot to bot interactions feel, to me, a lot more human.

Katie Rose Pipkin

Do people ever wonder about you, the human behind the bot?

I think many people that follow Tiny Star Fields do not understand that it is a bot! Or that bots are even on Twitter. The predominant interaction that seems to occur runs along the lines of “DO YOU SLEEP” or “what is this” or “i love these thank you so much for making them all the time”. I find that disconnect pretty delightful- the assumption of a (very) dedicated human somewhere. I’m also fond of the interpersonal conversations that happen in the comments, often having nothing to do with the original stars at all- it seems to occasionally function as a bit of a forum for strangers to connect.

Where do you see Twitter bots, or social media bots in general, evolving?

I have found myself moving off of Twitter and back into non-social spaces for a lot of my work. Part of this is probably personal; my interests shift project-to-project. Part of it is intrinsic limitations in the media, the 140 character limit, the difficulty of keeping up with Twitter’s often evolving terms of service. I am interested in physical robots, or the housing of digital spaces- where these bots actually live- and a lot of my studio practice is in exploring actual tangible machines right now. Some of the best bots I’m seeing out of others use neural nets, or very clever source material. In my own work, I am looking forward to more physical-digital integration, especially as I pick up some new toolsets required for more complicated work. I have an interest in biological emulation and in the hidden data that Twitter links to every tweet (perhaps my next bot will not be readable on the Twitter webclient, but instead comes alive in an API call?).

There is also a small part of me that feels like others have taken up the call (and doing it better than I ever could have). This is to say, Twitter bots are in a kind of renaissance- tools like George Buckenham’s Cheap Bots Done Quick (which uses Kate Compton’s Tracery) and the plethora of tutorials and frameworks have radically democratized the process, and it seems like every day I see someone new to this space building interesting or beautiful things. I am learning as much from newcomers to the form than anything!

In short, for the future- who knows? But at the moment, bots are serving as a fascinating space to test out new ideas, construct entity and artwork of generated text and data, and publish those experiments to an audience who are excited to see them in the world. What more could one hope for?

Finally, what are your favorite bots at the moment?

https://twitter.com/CreatureList – automata bestiary from @samteebee

https://twitter.com/FFD8FFDB – image-processed security cameras by @derekarnold

https://twitter.com/imgconvos – a @thricedotted answer to image-bot loops

https://twitter.com/everycolorbot – the first bot truly dear to me still going strong, thanks to @vogon

https://twitter.com/reverseocr – a @tinysubversions bot that randomly draws until it hits whatever word it is trying to match in an OCR library

https://twitter.com/ARealRiver – the only real way to view this (very clever) bot is in its own timeline, probably on mobile. from @muffinista

https://twitter.com/LSystemBot – l systems by @objelisks

https://twitter.com/INTERESTING_JPG – a bot-form of deep learning, which attempts to describe human images with computer vision, by @cmyr

https://twitter.com/park_your_car – compelling use of google maps highlighting available car-space by @elibrody

https://twitter.com/wikishoutouts – shoutouts to the disambiguation pages of wikipedia

https://twitter.com/soft_focuses – a very quiet mysterious bot from @thricedotted

https://twitter.com/TVCommentBot  –  attemped image recognition of television, @DavidLublin

https://twitter.com/GenerateACat – procedural cats – @mousefountain and @bzgeb

https://twitter.com/pentametron – a bot that looks for tweets in accidental iambic pentameter by @ranjit

https://twitter.com/RestroomGender – @lichlike’s gendered restroom sign generator

https://twitter.com/digital_henge  – this bot by @alicemazzy tweets moon phases, eclipses, and other solar and lunar phenomena

https://twitter.com/a_lovely_cloud – digital cloud watching from @rainshapes

https://twitter.com/the_ephemerides – computer generated poetry with outer space probe imagery, @aparrish


Taina Bucher is an Assistant Professor at the Centre for Communication and Computing, University of Copenhagen. Her research and teaching focuses on the intersection between media theory, software studies and social media. She holds a PhD in media studies from the University of Oslo (2012), on a dissertation about the ways in which software influences sociality in social networking sites. Besides her academic writing, she regularly publishes interviews with media artists and features on social media art.

To find out more about Katie Rose Pipkin’s latest projects, please visit http://katierosepipkin.com