I, Twitter Bot

00:17:16 minutes

As of 2014, Twitter estimated that as many as 23 million of its 271 million users were bots, or automated accounts taught by code to Tweet according to the desires of their creators. For example, you might see that one bot posts a photo from NASA’s Cassini mission every hour, while others recycle politicians’ old tweets to generate new ones that sound just as believable.

Bot writers Darius Kazemi and Kate Compton talk about the bots that are writing poetry, dreaming up New York Times thinkpieces, and even helping activists stay motivated and educate their followers. They share what they think bots can do for Internet users, and why they suggest that more people try creating them.

We include some examples of Twitter’s bot variety below:


Uses past tweets from @SciFri to generate tweets that sound…almost…like they could be from @SciFri, but a little weird. A common method for creating bots.


Rewriting the same 10 seconds of Disney’s “Gaston” with new sets of rhyming lyrics. Started as just text, but now includes computer generated singing as well.


Images of Saturn and its moons from Cassini. From the people who brought you @LandsatBot (Earth) and @HiRISEBot (Mars).


Fake thinkpiece headlines generated from topics in real thinkpiece headlines. Inspired by all those thinkpieces about millennials.


Maps of fantasy worlds, complete with city names and coastlines. Code uses rules on how water behaves in order to sculpt realistic geography. Built by a glaciologist.


Posts random open access images from New York’s Metropolitan Museum of Art’s collection. Friends include @MOMARobot and @TateBot.


Random images from outer planet space probes paired with computer-written poetry. Creator, @AParrish, says she sees similarities between both the probes and the poetry: “Both space probes and generative poetry programs venture into realms inhospitable to human survival and send back telemetry telling us what is found there.”


Education and activism tool for Black Lives Matter. Sends encouraging messages to activists besieged by trolls, tells users how to contact their elected representatives about legislation, and more.


Generating examples of what the NSA may have on us from their data collection programs. Pairs random names, subject matter, and actions with real examples of companies that have given data to the NSA.


Tiny stories about the real people in Census data. Each tweet is a row of data, rewritten to describe the person it represents.

Segment Guests

Darius Kazemi

Darius Kazemi is an internet artist and programmer, and the co-founder of Feel Train.  He’s based in Portland, Oregon.

Kate Compton

Kate Compton is a programmer and a computer science PhD candidate at UC Santa Cruz, in Santa Cruz, California.

Segment Transcript

IRA FLATOW: This is Science Friday. I am Ira Flatow. Twitter is the home to hundreds of millions of users but the company says that tens of millions of them, more than 23 million at last count in 2014, may actually be bots, not humans at all. These Twitter bots are automated pieces of computer code programmed to react to you and to other folks or they just sit there and tweet their code-generated content.

Many of the original Twitter bots may have been designed to spam, to send out junk. But the bots have multiplied and so have their missions. Many you see today are quirky. Some are entertaining and even artful.

Here’s what I mean. Take the bot called “Botston.” This bot takes the song “Gaston” from Disney’s Beauty and the Beast and creates new bizarre rhyming lyrics to the music.

BOTSTON: [SINGING IN ROBOT VOICE] No one thinks like Gaston or lights [INAUDIBLE] like Gaston, plans to [INAUDIBLE] for a run down the notes like Gaston.

IRA FLATOW: That’s a bot called ThinkP– there’s a bot called “ThinkPieceBot” that mashes up common headlines to create a tweet that might look like, “Is this big data to blame for swine flu?” Well, is it? Another just tweets variations of the scream “ahh,” variations of that on a regular basis. MuseumBot tweets photos from the Metropolitan Museum of Art’s collection.

Well, if it’s not your cup of tea, you better get used to it because the world of bots is not only weird but wide and growing. We even made our own bot and you can see some of the stuff it’s tweeting at sciencefriday.com/twitter plus more bot examples on our website at sciencefriday.com/bots. They’re all there, both of them.

And what’s fueling this craze? Let’s find out. Darius Kazemi is a digital artist and programmer, founder of the annual Bot Summit conference and co-founder of the company Feel Train. He’s written dozens of bots for Twitter and other platforms. Kate Compton is a PhD candidate at the University of California, Santa Cruz. She’s a bot writer and creator of one tool that helps non-programmers make their own bots.

Welcome to Science Friday.

KATE COMPTON: Hi, I’m pleased to be here.

DARIUS KAZEMI: Glad to be here.

IRA FLATOW: Nice to have both of you. Darius, besides writing bots, you’ve done a lot of work organizing the bot community and you’re sort of known as the “bot dad” and the “bot father” in bot writing circles. You have a summit that brings together bot makers. What’s so great about bots and what makes them great?

DARIUS KAZEMI: I like to think of it as keeping the internet weird. I remember the internet from 10, 15, 20 years ago and just how strange it was and how everything seemed like a new, unexpected thing to me. And I don’t know. I think over time, things get more corporate and curated. And I like that bots just throw unexpected things into your internet experience.

IRA FLATOW: So what are some of your favorite bots that you’ve made?

DARIUS KAZEMI: Well, that I’ve made?


DARIUS KAZEMI: I really like Two Headlines, which it basically takes one headline from the news and then swaps out a different source for it– excuse me, subject for it. So for example, it might take a headline like “Miley Cyrus to Entertain Millions at a Concert” and then it would change that to “Vladimir Putin to Entertain Millions at a Concert.”


IRA FLATOW: Darius, our digital producer Daniel Peterschmidt made us our own Twitter bot. We named it “SciFri_ebooks.” Our Twitter bot takes language from our old tweets and rehashes it into new, almost believable program ideas like “Someday, You May Unlock Your Future Phone with a Grain of Salt.” This is kind of an approach to bots use, isn’t it?

DARIUS KAZEMI: Yeah. Specifically, there’s a whole genre of bot known as the “ebooks bot,” sort of named after a famous– it’s hard to explain but it was a bot that was not a bot in the end. It was called “Horse ebooks” and it was putting strange, amusing, spam-like things. Basically, it turned out that it was a bot for a while. And then after a while, a human took it over, a human poet, an artist, and started tweeting from it. And that’s when it got really popular and then people found out and felt betrayed that it was written by a person rather than a bot.

But generally speaking, the ebooks bots, yeah, they’re a whole genre. I’m Tiny Subversions on Twitter and I might have a Tiny Subversions ebooks bot, for example, that would take all my old tweets and jumble them up using a statistical method known as a Markov process, a Markov chain, and stitch them back together into kind of a word soup that is somewhat grammatically plausible but not actually grammar most of the time.

IRA FLATOW: Darius, one of your first projects with the company was a Twitter bot called “Stay Woke Bot,” which was commissioned by the Black Lives Matter movement. What does Stay Woke Bot do?

DARIUS KAZEMI: Stay Woke, it’s a collaboration with DeRay Mckesson and his crew of people with the Movement for Black Lives. And what Stay Woke does is it’s meant to sort of take some of the load that online activists have to deal with all the time. What DeRay was finding was that he was getting kind of inundated with the same sorts of questions and sometimes harassment over and over. And so we designed this bot to help him kind of automate some of his responses to that. So he has a Google spreadsheet that he can just update with certain keywords and then have particular canned responses.

IRA FLATOW: Kate, you write bots, too. What are some of the ones you’ve made? Give us an idea.

KATE COMPTON: Let’s see. What have I made? I made– actually just recently, two days ago, I made an oblique strategies bot. Well, so there had already been Oblique Strategies Bot because it turns out Oblique Strategies is a card game about introducing randomness into your creative practice and so that’s inspired a lot of bots that use randomness as part of their creative practice.

So I made something called “Art Idea Bot” that gives you art ideas. I’ve made something that draws spaceships and then makes tiny little sci-fi stories next to them. And then I made something that just makes tartans all day long, different colors of tartans.

IRA FLATOW: Well, that brings up the question. What are you both– let me ask you, Kate, first. What gap in people’s lives do bots fill the need? What need do they have that bots fill?

KATE COMPTON: I like to think of them almost as a cuckoo clock, that it’s a little decorative object that you have in your life and you know that every hour or once a day or once a month, it’s going to, I guess rather literally, tweet at you something new and bizarre and surprising. And I think a lot of what makes bots so fun is that they have no filter. They’re randomness.

So much like Oblique Strategies, they sometimes put things together that you wouldn’t expect them to put together. Or for example, the Metropolitan Museum of Art bot tweets things from their collection but it doesn’t just tweet the good stuff. Sometimes, it’ll just be a shard of pottery or a scrap of fabric or a not very interesting vase. But for a moment, it takes this uninteresting thing and says, hey, look at this.

IRA FLATOW: Yeah. Kate, do you think everybody can write a bot of their own?

KATE COMPTON: Oh, absolutely. So the tool that was a collaboration between me and George Buckenham is called Cheap Bots, Done Quick! And if you have a Twitter bot or if you have a Twitter account, you can log in there and make a bot in about five minutes. The challenge is really just thinking what you want to bot about.

You can make something that wishes your mom happy birthday. You can make something that writes poetry. You can make something that makes art. You can make something that comes up with new Pokemon. Really, there’s kind of no limit, in the same way that there’s no limit to what you could write a poem or a short story about. You can bot about just about anything.

IRA FLATOW: Darius, is there any way to know if you’re talking to a bot versus a real person on Twitter?

DARIUS KAZEMI: Well, yeah. The first thing, of course, is just to check the profile to see if it already self-identifies as a bot. Most of the bots that Kate and I have been talking about do identify as bots. It’s actually really important for something that’s like a work of automated art to sort of self-declare as a bot because they do put together these strange combinations of things. And there was one bot creator whose bot accidentally put a death threat together and ended up being visited by the police in– I believe it was in the Netherlands.

And so declaring that something is a bot can be important for both aesthetic and practical reasons. But in terms of, oh, this person– this account just followed me on Twitter. Can I tell if it’s a bot or not? Typically, what I do is I look at– especially if you’re thinking about spam, look and see if there are a lot of other bots that are followed by it or if it seems to have a– not just look at one or two things that it’s done but look at its entire history and see if it’s somewhat human or if you can start to see a pattern in it. That said, I’ve found actual people, live humans, whose Twitter accounts I was positive were bots.


IRA FLATOW: Kate, do you think that bots are going to evolve to learn, for example, to talk to one another? We’re going to have tens of millions of bots communicating?

KATE COMPTON: A couple of them already do. I guess the big question is, how do bots listen? And a lot of bots, much like humans, will listen for when you stop talking and then will say what they wanted to say, anyway. So the endless screaming bot that you mentioned earlier, which is one of my all-time favorite bots– I don’t know why I love it so much but I do– it will respond if you tweet at it.

So I can say, endless screaming, how are you doing today? Ahh. Oh, that’s great. How is that working out for you? Ahh. So you can already communicate with bots and they’ll just play the straight man, kind of responding however they wanted.

But then there’s also bots that– I’ve seen this mostly with the art bots that art bots will respond to each other. So there’s the MuseumBot that tweets out museum works but there’s also an AppreciationBot which then responds to what the MuseumBot has posted with a little generative appreciation of it.

DARIUS KAZEMI: Right. And I didn’t build that for MuseumBot. That was another bot maker, Michael Cook, who put that together.

IRA FLATOW: Hmm. So is it possible to make bots that could help you health-wise, remind you maybe to take your medicine, remind you to go to the gym, things like that?

KATE COMPTON: Let’s see. I guess there’s Nora Reed’s– is it Drink Water Hydration Bot, I think–

DARIUS KAZEMI: Yeah, Hydration Bot–

KATE COMPTON: That periodically reminds you to drink some water. Yeah, it’ll be interesting to see when we get more people making these bots. A lot of these Twitter bots are kind of meant for large broadcasts, like I’m trying to tweet at everyone.

Is there something that can just tweet at me? If I’m the only person following my go for a walk bot and it just reminds me to go for a walk or “it’s time to feed your cats, Kate.” Yeah, there’s no reason that all bots have to be for the entire world. Maybe they can just be for one person.

DARIUS KAZEMI: There was a great British bot called Bin Night, which is garbage night in the UK, and it just reminded you to take out the garbage on garbage night. I thought that was great.

IRA FLATOW: I could use that one. And the bots are not limited just to Twitter, are they?

DARIUS KAZEMI: No, that’s correct. I think a bot– in my experience, a bot can live anywhere that there’s technology for humans to talk to other humans. I think a robo dialer that calls you with a survey on the phone is a bot. It’s just a bot that’s happening on the telephone system instead of on Twitter.

IRA FLATOW: How exactly do they work? How complicated is it to code? Kate, if you’ve got a tool that I can use, it’s got to be not very complicated at all.

DARIUS KAZEMI: Sure, it’s almost like– well, so there are many different approaches and Darius has a lot of his own approaches. Mine is more like Mad Libs, where you give it a recipe for something you want to generate and then you give it things that can slot into that recipe.

So if you wanted to make mixed drinks, a little bot that makes mixed drinks recipes– you say a drink is made out of two liquids and a glass. And a liquid can be orange juice or soda or all these different things and a glass can be a bucket or a cup or mug or a tiki glass and that will be a bot. So it’s really just kind of filling in the components of your world and telling it a recipe for how to combine them.

IRA FLATOW: We’ve got a lot of folks interested in the bots. Let me remind everybody that this is Science Friday from PRI, Public Radio International. I’m Ira Flatow talking with Darius Kazemi and Kate Compton.

A couple of tweets– I hope they’re not bots– that came in from Techno Terrorist says, “God Tribute is one of the weirdest on Twitter.” And maybe you’ve heard of that one. A lot of people are coming in, people saying that they love to have the bots. And what percentage do you think– are bots going to grow so that they take over a large percentage, if not already, of the tweets that are going out there?

DARIUS KAZEMI: Well, if you look at some of Twitter’s SCC filings, buried in there was the admission that some very large percentage of Twitter accounts are bots. And part of that is because Twitter is somewhat lenient about bots. There are other platforms that try to shut down bots– put a lot of effort into that. So really, a lot of the future of bots and where it’s going, where they go, depends on these platforms and whether they embrace the bots or not.

IRA FLATOW: Wait. I’m sorry. Kate, did you want to–

KATE COMPTON: Oh, yeah. Just I have a fashion bot that I’ve been working on, trying to make the first-ever bot-made fashion magazine and I really wanted to have it on Instagram but Instagram has a very strict no bots policy. So Instagram is at least nominally bot-free at the moment.

IRA FLATOW: Well, that raises a question of whether you have any kind of fear that these bots could run amok, out of control.

DARIUS KAZEMI: There’s a limited amount of running amok that a purely software bot that can just tweet museum pictures and stuff can do. But I certainly put safeguards in a lot of my bots. I have a language filter that I apply to all of my bots that just basically as a rule of thumb, it’s words that I wouldn’t say I don’t want my bots to say. And so they’re not necessarily family-friendly but there are definitely some things, like racial slurs and stuff, that I just don’t want the bots doing.

And so I apply that uniformly across all of my bots. There’s also just you have to program these things. I think it’s important to program these things so they don’t annoy people or spam people, at least people who haven’t signed up for it already. So I could make a bot that floods your feed with cat pictures every 30 seconds or something. But if I did that, I would make it be purely opt-in, like you would have to consent to getting that flood.

IRA FLATOW: Kate, let’s talk about your work on the online artist colony for bots.

KATE COMPTON: Oh, yeah. So this is very new, very novel work. It’s still very much in production. But the idea is that I make these art bots. So this is with a collaborator, Jonathan [INAUDIBLE]. And we have basically a bot art school that produces art bots. Each art bot has things that it likes and things that it’s able to paint.

And then these little art bots can communicate with each other. They can hate their own work. They can love somebody else’s work. They can look at somebody else’s work, wish that they could make that, take apart their work or the other bot’s work, and learn new ways to make red rectangles or yellow triangles or whatever it was that their personal little bot minds wanted to be able to make all along.

IRA FLATOW: How does a bot look at someone’s artwork?

KATE COMPTON: It kind of goes back to the, how do bots listen or how do bots look? This is just a toy program for the moment. It’s very new and so we’re just really looking at color. But also there’s an algorithm that detects cat pictures. It will look at a JPEG and will try to see if it’s a cat or not. So we’re hoping that one of our bots will really just wish that it could paint cat pictures all day and will collect artwork that looks the most like cats.

IRA FLATOW: Well, this is fascinating. I’ve got to tell you, I didn’t know very much about bots until we did this program today. But I’ve learned an awful lot and I’m going to start looking for them myself. I thank you both for taking time to be with us today and good luck with your bots.

KATE COMPTON: Thank you.

DARIUS KAZEMI: Thanks for having me.

IRA FLATOW: And Darius Kazemi is a digital artist and programmer, a bot writer, and co-founder of Feel Train. Kate Compton is a bot writer and computer science PhD candidate at UC Santa Cruz.

Copyright © 2016 Science Friday Initiative. All rights reserved. Science Friday transcripts are produced on a tight deadline by 3Play Media. Fidelity to the original aired/published audio or video file might vary, and text might be updated or amended in the future. For the authoritative record of ScienceFriday’s programming, please visit the original aired/published recording. For terms of use and more information, visit our policies pages at http://www.sciencefriday.com/about/policies

Meet the Producer

About Christie Taylor

Christie Taylor was a producer for Science Friday. Her days involved diligent research, too many phone calls for an introvert, and asking scientists if they have any audio of that narwhal heartbeat.

Explore More