Why Weird Twitter

tl;dr There’s no such thing as Weird Twitter

Will there be a mythology in the future, they used to ask, after all has become science? Will high deeds be told in epic, or only in computer code?

And after the questing spirit had gone into overdrive during the early Space Decades, after the great Captains had appeared, there did grow up a mythos through which to view the deeds. This myth filter was necessary. The ship logs could not tell it rightly nor could any flatfooted prose. And the deeds were too bright to be viewed direct. They could only be sung by a bard gone blind from viewing suns that were suns.

R.A. Lafferty, Space Chantey

Imagine that there is a community or culture of people that use social media–let’s focus on Twitter–in a particularly interesting or funny or outlandish way. Would you give it a name? Would you try to understand its size or its structure? Its history? Its purpose? How would you go about doing that?

Could it be studied by an anthropologist? A data scientist? An economist? A philosopher? A critic? A journalist? Could it ever understand itself?

I’m getting ahead of myself. Let’s start with a name: Weird Twitter.


This is a Google Trends query for the string “weird twitter”.

There have been two spikes in search popularity of “weird twitter” in the past year. The second spike corresponds to the publication of the brilliant “Weird Twitter: The Oral History“, by Herrman and Notopoulos, (BuzzFeed, April 2013). The authors explain:

Weird Twitter is vast and amorphous; what it looks like depends hugely on whom you follow, when you followed them, and what you find funny. … Some of its best writers have a few hundred followers, while others have tens of thousands. Styles of tweeting and types of jokes that originated among its small sects have bled out into the mainstream: Even to comedians, these are some of the funniest people on Twitter.

The first spike in October 2012, which marks the beginning of a consistent hum of search activity for “weird twitter”, coincides with this tweet by erstwhile pseudonymous Gawker writer Mebute Sese Seko.

At the time, Mebuto Sese Seko had almost 10,000 followers. This tweet triggered a series of events that lead to the wider adoption of the term.

Notably, Mebuto didn’t use the phrase “weird twitter” himself, and he linked to screenshots, not to web pages. The second image was of an anonymous Quora post that identified Weird Twitter as Twitter’s equivalent of 4chan’s /b/, or “Random”, subcommunity. The first image pictured a blog post I had written the previous summer.

Boundaried symbolic network community

A lot was going through my head when I wrote that blog post. In the Spring of 2012 I was reading Anthony Cohen‘s work on the symbolic construction of community. For Cohen, a community is constituted by its creation and use of symbols. Especially critical for the community’s identity are the symbols it uses to mark its boundary–members and non-members. I was also reading Caroline Haythornthwaithe‘s work on Social Networks and On-line Community, which emphasizes the topological structure of on-line social networks over symbolic meaning-making.

Schematic diagrams of Cohen and Haythornthwaithe's models of community.

(I did this preliminary work in on-line community detection in collaboration with a classmate, Dave Tomcik.)

At the same time I was following a few of what now would be called Weird Twitter accounts. It got me thinking: most work on virtual communities in cyberspace depends on technical infrastructure to provide the community boundaries. A mailing list is a community circumscribed by the technicalities of mailing list membership. A web service like Reddit supports multiple communities by supporting multiple distinct subreddits. But Twitter supports the growth of an ad hoc network structure without distinct watering holes demarcated in the user interface.

How could one identify a community within such a social network? I had a hunch that these networks, which would depend on the organic social connections between individuals and not the commercially built and sustained technical environment, would be special.

Schematic diagram of a boundaried symbolic networks and containment relations between its lexicons

What sort of digital signature would such a community show? According to my reading of Cohen, it would be involved in vigorous, complex dialog about, among other things, what symbols to use to represent itself. Since the digital environment is one in which symbols (e.g. words) are constantly flying around and being recorded, I thought this kind of community could be algorithmically detected. Which I find both thrilling and chilling. Eric Snowden’s recent whistleblowing has let America know its on-line activity is under state surveillance all the time, on top of the surveillance by commercial interests.

Chaotic resistance

Duchamp's FountainI believe we have an opportunity to use the wealth of data available now to really advance social science. But the reality is our research will, if successful, be used for political manipulation, commercial advertising, and other kinds of social manipulation and control. For me, this makes it imperative that I present my research to the public. If I work on methods for on-line community detection, I should try to make those insights usable by communities to understand themselves and evade detection if they desire.

Most rhetoric about evading on-line detection is about making less information available. That makes sense for the individual. But in aggregate, this cleans the data set, making it easier to find patterns that haven’t been self-censored away.

The biggest challenges to behavioral data scientists are not the availability of data, but data’s complexity. If data has a high Kolmogorov complexity and perhaps logical depth, it will be very difficult to extract patterns from.

In other words: you cannot master noise and chaos. It is the abyss staring back. Much as Dadaism was a tool for eroding the establishment and Situationists sought to challenge society through liberated, authentic expression, social media users can resist surveillance by making their interactions more wild, original and complex. Big Brother is watching, but he can be blinded by confusion fu.

We are always already watched over by machines of love and grace, and other kindsI thought I had found a nexus of this kind of noisy on-line behavior on Twitter. If what I was seeing was a community at all, it was a community of chaos and exploration. And it knew it. I had read the term “weird twitter” in @regisl‘s tweets in March 2012 when he was writing a lot of exploratory, reflective thoughts on Twitter culture. His and others insights into virtual community were profound, echoes of some of the earliest musings on virtual community, such as by Electronic Frontier Foundation founder John Perry Barlow in “Crime and Puzzlement” (1990):

As a result of [the opening of Cyberspace], humanity is now undergoing the most profound transformation of its history. Coming into the Virtual World, we inhabit Information. Indeed, we become Information. Thought is embodied and the Flesh is made Word. It’s weird as hell.

I figured that holding a mirror up to the noise, crudely describing and interpreting practices that could not be interpreted or described, could only make the chaotic system more complex. At the same time, it would be a test of Cohen’s theory of the symbolic construction of communities in an on-line context: what happens when a community confronts a symbol, in this case the string “Weird Twitter”, that purports to mark its boundary?

Stuart Geiger had introduced me to M.C. Burton’s idea that “trolling is the new critique.” I was interested in trolling as an experimental method. A grown-up Internet kid who had done and been dished his share of trolling, I figured it was time to put those skills to good use: clumsy live field notes.

The offending post

According to the Encyclopedia Dramatica, this kind of trolling is a Philosopher Attack (“a type of flame war
where a terminally bored, yet well educated person or group ambush an innocent bystander or group, who were just minding their own business”). I posted in August. Nothing happened.

Then Jeb Lund tweeted about it.

Attention economy

I was wrong about Weird Twitter.

One way I was wrong was that I had imagined Weird Twitter to be a kind of complex utopia untainted by the capitalist attention economy. I had been following a small sample of what I had perceived to be a minute and innocent avant-garde art movement.

While there was a kernel of truth to that naivete, I hadn’t understood that the phenomenon I was calling out was much larger than I thought and always-already in tension with the web gossip press. I also hadn’t thought about the phenomenon I was pointing to in terms of pseudonymity, doxing, secrecy, and privacy.

Mebutu Sese Seko is not Jeb LundNobody represents the intersection of the pseudonymous web and the web gossip press better than Jeb Lund, alias Mebute Sese Seko, who more than anyone else is responsible for the Weird Twitter meme because of a single tweet disparaging nerds. At the time he posted that tweet, Lund was writing for Gawker and still pseudonymous.

I think he did it on purpose.

We know a lot more about Jeb Lund than we did in October 2012, when most people who had heard of him thought he was a Brooklyn hipster and just possibly black. In February 2013, Lund doxed himself, explaining that he’s a 35-year old white guy from Florida who in past Internet lives had been a pseudonymous message board troll. Twelve years ago, his real name had been discovered and, as a prank, used in messages supporting some vile causes. As a result he received threats, suffered emotional distress, and deleted his Internet history. He reemerged under the name of a brutal African dictator writing an opinion column for Gawker, which in 2012 seemed to specialize in disclosing private information and exposing high-profile pseudonymous web personalities. In December, Drew Johnson, a blogger frustrated with Gawker’s role as on-line “morality police”, exposed Mebute Sese Seko as Lund. This perhaps prompted Lund’s later self-disclosure and retirement from Gawker.

When Stewart Brand said “Information wants to be free,” he may not have had in mind the fragility of pseudonymity in social media. But just as the second law of thermodynamics says that entropy in a closed system never decreases, so too do our genies of personal identity never get back into their bottles. So too, perhaps, with an on-line community that had avoided identity by not being named.

But was it truly only a matter of time? As is clear from the trend chart, the linked blog post did not trigger the information cascade that brought “Weird Twitter” into the public eye. In another world, it might have been obscure forever. So why trigger the cascade?

credit Hipster RunoffDespite our often inadvertently public lives and data, what saves us from constant exposure is scarcity of attention. This scarcity drives the market for on-line advertising and on-line journalism. Or, to put it less politely, content farming–the “generat[ion of] large amounts of textual content which is specifically designed to satisfy algorithms for maximal retrieval by automated search engines. Their main goal is to generate advertising revenue through attracting reader page views as first exposed in the context of social spam.”

In “Lana & Me: Our Dark, Abusive, Co-Dependent Relationship on the Content Farm” (Hipster Runoff, January, 2012), Carles explains the existential-crisis-inducing problems of being a professional indie music blogger in a hyperquantified world driven by page views.

I have a blog called HIPSTER RUNOFF. Every day, I wake up, open my laptop, and type words that are stored in the internet as ‘content.’ My goal is to ‘get as many hits’ as possible because I metaphorically ‘have mouths to feed.’ I realize that at this point, it doesn’t matter if my content is ‘premium’, pseudo-brilliantly written web_prose or just ‘link-bait-wave,’ I was fortunate enough to not have gotten lost in the ‘long tail’ of indie music + Gen-Y-opinion-driven coverage blogs. Every day, I prey upon different buzz topics, exploiting my voice, but more importantly, my position as a ‘recognized outlet 4 buzz’ to try to trick people into thinking I am ‘relevant’, which basically just means that I am trying to make ppl talk abt my blog and get them addicted to my web brand even if they hate it because even when they are like ‘OMG THAT’S TOTAL BULLSHIT’ it is just some sort of post-grassroots-h8-wave-warketing.

Carles points to the rising media tide of commentary of controversially inauthentic pop singer Lana Del Rey as a point where indie music blogs converted from tastemakers to content farmers.

LDR should probably make other alt famous humans and upcoming buzz humans incredibly nervous, because the silent agreement of ‘pandering’ to indie celebs is over. The concept Indie celebrity may or may not exist, but a successful blog now must realize that the way indie blogs ‘baby’ the ‘alt famous’ in order to preserve relationships can no longer exist. Mudslinging will be beneficial to both parties. Maybe after LDR, it will no longer be a good idea to be ‘an anonymous’ project because then dumb blogs will just use Google to construct an unfair narrative for you.

In particular, Carles showers accolades on Gawker’s founder Nick Denton, “the guy who invented Gawker and revolutionized modern content farming,” lamenting “As I struggle to deal with my own periodic content farm existential crises, I wish I could have just invented Gawker. We could’ve had it all. Rolling in the deep pool of unique visitors.”

Around this time, Gawker began its regular postings on Weird Internets. Strange things happening on the Internet were newsworthy. That included events that took place entirely on Twitter. Engaged in a kind of legitimacy exchange with the Internet’s absurd, by acknowledging them as newsworthy it built them up as an audience which in turn would consider it, if not a credible journalistic outlet, at least worth paying attention to.

So when Lund tweeted the controversial links, some were not surprised at all.

If the plan was to immediately capitalize on the new “Weird Twitter” sensationalism, it backfired. Gawker was quickly singled out as the most likely media outlet to exploit it for page views. To have gone ahead and done it would have been egregiously uncool. Complicating the dynamics of economy and credibility, the web commentators who would write such an article saw themselves as part of the craze.

The difficult task of providing the web press take on Weird Twitter fell to BuzzFeed writers Herrman and Notopoulos. In a refreshing change from Buzzfeed‘s lampooned listicles, they wrote with the ethical sensitivity of ethnographers, not journalists — their article consists mostly of quotes from conversations with high-profile community members. In the words of Carles, it was an opportunity for “alt famous humans and upcoming buzz humans” to construct a narrative of themselves before anybody (else) did it unfairly. In so doing, they shifted valuable attention onto a select social circle. Despite their care in writing the article, the authors report a loss of social capital.

Shortly afterward, in the wake of the Boston bombings, Herrman and Smith wrote for Buzzfeed of a rediscovered journalistic ethic that synthesizes the “mouths to feed” imperative of topicality with a genuine sense of civic duty:

The media’s new and unfamiliar job is to provide a framework for understanding the wild, unvetted, and incredibly intoxicating information that its audience will inevitably see — not to ignore it. A Reddit post seen by millions without context is worse for the story, and the public, and to the mission of reporting than the same post in a helpful and informed context seen by many more. Reporting is no longer a question of whether or not to dignify new and questionable information with attention — it’s about predicting which of it will influence the story, and explaining, debunking, or contextualizing it the best we can. That is, incidentally, what our readers want.

When I began experimenting with Weird Twitter, I didn’t understand that my Petri dish was the fuzzy and paradoxical frontier of the attention economy. Now, I’d say that Weird Twitter is at least 45% coconstructed by the media. The ongoing microcosmic task of mining it for memes–“extricat[ing] concepts from Weird Twitter, but on a macro-scale”–has been claimed by vrunt, editor of up-and-coming BuzzFeed rival FeedBuzz.

But back to science. Because the trajectory of “Weird Twitter” between Lund’s tweet and Buzzfeed‘s “Oral History” gives us a choice opportunity to study the diffusion of a meme in the wild.

Observer effects

Lund’s tweet spurred a “collective eye-rolling” at and within Weird Twitter; some users contacted me to complain. Since the initial reactions suggested I had a lot to learn about Weird Twitter, I began collecting data through blog posts that might be called cultural probes.

Explanation of backscatter analysis
One trick that scientific researchers have been using for a long time is explaining data after the fact as if they had planned it all along as a well-structured hypothesis and experiment. Whatever my original intentions were as I wrote those probes (which I’ll admit I’m a bit fuzzy on–I recall writing under the influence of a tall glass of scotch), I can retroactively and definitely say I was conducting a complex set of experiments-within-experiments.

Georg LukacsAs I’ve explained elsewhere, one purpose of this follow-up post was to get an approximate measurement of the size of Weird Twitter. This technique was inspired by Vern Paxson‘s work on backscatter analysis of Internet activity. As the link to the second post spread through Twitter, I was able to observe traffic to an otherwise “dark” part of the Internet (my blog), which in turn gave an indirect measure of the size of the meme infected cluster. (It got 2.9k hits in the first three hours after it was posted, which I take to be an approximate lower bound on the community’s size at that time.)

The irony of this technique is that I literally didn’t know who I was talking about. Many people were upset by the probe because they interpreted it as being about them. But in having that interpretation and acting on it, they made it so. Matt Sweeney, a lead data scientist at Uber, comments that this is analogous to the observer effect in physics:

I think this is doubly interesting in this case, where there aren’t *any* rules or guidelines or even a clear idea who is a part of it. In a vacuum like that, commentary could have a very real effect on the subject matter. Everyone is projecting meaning, participants and observers alike.

Would this otherwise nameless community be “symbolically constructed” as Cohen’s theory suggests? Put another way, would the name “Weird Twitter” stick? And would the community become more cohesive if it were iteratively called out as a ‘real thing’, whether or not it was one?

Soon enough I wasn’t the only one reifying Weird Twitter. First within the community itself and then from without, Weird Twitter autoreified in a process that culminated in the “Oral History” article mentioned above.

How did this happen?

The clue might be in Rob Horning’s “Fragments on Microcelebrity” (The New Inquiry, October, 2012) Horning asks if some of the problems of fame “once reserved for reflexive and narcissistic rock stars, now potentially afflict us all”:

Does microfame yield macro shame? My experience with Facebook has been double-edged in that way: It seemed I had a chance to redeem all that time I felt ignored. But I only rediscovered the same disgust with myself for wallowing in that miserable egomania, as Facebook forced me to recognize yet again that my tastes and experiences aren’t really my own, because I still want so badly for people to applaud me for them.

Micro-celebrities, just like indie music blog writers, are trapped in a double bind between fame and authenticity. And many of the most influential members of Weird Twitter are micro-celebrities. Perhaps what happened was this: these influentials were given the opportunity to draw attention to themselves by adopting a external-facing name, but had the discomfort of knowing that the label was artificial and self-promoting, As a result, Weird Twitter was victim (and beneficiary) of its own inward-facing Streisand effect as it attempted to discourage use of the term through chiding subtweets (“referring to a particular person [or thing] without directly mentioning them”). This sensitivity resulted in Weird Twitter being a percolating vulnerable cluster for the meme describing itself.

Cover art of The House of Inconstant Horses, by Theron JacobsThat’s one way of explaining what happened. But there is evidence that this misrepresents the sensibility and self-consciousness of Weird Twitter. In what may be an emic allegory, Theron Jacobs (@TPHD) has written The House of Inconstant Horses, a sequence of micro-fiction, on his Tumblr. He writes:

“I am a migratory donkey!” said the donkey.

The horse looked superior.

“Donkeys don’t migrate,” explained the superior horse.

“And you,” said the donkey “are wearing a hat!” in such a way that seemed as if he had won the argument and was having a very good time all at once.

Several horses snickered.

The hatted horse grew furious.

“I’ll ha—” but he was interrupted by the migratory donkey.

“And you look quite excellent in it too,” crooned the donkey in a most winning manner, “but the fact is that horses do not wear hats!”

And here several horses gasped.

Some of the gasps were sincere and some were for the sake of deepening the theatre of the moment—it is important to remember that inconstant horses are tricky creatures, and it is often impossible to tell when they are earnestly involved or when they are simply playing along to amuse themselves.

“In fact,” said the donkey, who by now had their complete attention, “you’re all wearing hats.”

And they were! The horses began, quite naturally, to grow alarmed. Was this little donkey a wizard?

He had enough sass to be a wizard. And it was certain that at least SOME horses were not wearing hats, and then he said we were wearing hats and we were. Magic!

“Pardon me,” begged the impetuous horse who’d first addressed the donkey and then left to go stand far away.

Whatever happened, by the time web business press outlet Business Insider covered Weird Twitter in December 2012, the meme outbreak had gone beyond damage control.

Research Ethics

In retrospect, I have real ethical qualms about this experimental method. There are important restrictions on what is allowed in human subjects research which involve being sensitive to subjects’ privacy, especially in the case of vulnerable populations like the juvenile and mentally ill. In my work on Weird Twitter, I was called out for violating an open secret and alerted to the presence of vulnerable members in the community.

When I looked into it, I discovered that university policy does not consider research into publicly available data to be human subjects research. But there was nothing unhuman about the data I was looking at (well, except for the data from the bots). And though it was public data, I realize now that by blogging I was violating the context in which the data had been created. Helen Nissenbaum has argued that privacy is a matter of preserved context, not just an issue of public versus private. Under this theory, I violated the privacy of Weird Twitter.

A feedback loopOn the other hand, I didn’t do it alone. In the flurry of activity surrounding these events, there was a rapid feedback loop between others’ observational writings on Weird Twitter and other web content aggregation communities. For example, Nick Douglas’s Slacktory blog post, “Weird Twitter explained“, was reposted to the forum Metafilter, where responses ranged from the unimpressed (“This seems like normal Twitter to me. I didn’t know it was weird.”) to reproachful (“Self-congratulatory masturbation. And I don’t mean the people tweeting the funny, I mean the author of this vacuous article.”). While elsewhere in the thread community members identified some of their own as part of Weird Twitter and went so far as to list off notable participants, Douglas was goaded into defending himself on the grounds that the community of interest was too insular to be recognized externally.

Also some people think I was maybe exploitative? Nnnnope. This article is more successful than the average Slacktory post, but it’s still only gonna hit maybe 10K views. Most of those are from people already interested in this subculture, whose biggest response will be to follow a couple of these users. No one’s getting rich, no one’s hurting anyone.

If there was a privacy violation, it was a group effort by those sharing links and discussing it. Luciano Floridi argues that in the Information Society we are faced with distributed responsibility and therefore need a distributed ethics. Maybe in an era of what Manuel Castells calls mass self-communication, unwinding the cyclone of distributed disclosure and scandal politics requires a collective effort at discretion.

Partly due to ethical concerns and partly because I didn’t want to interfere with the progress of the social reaction once it seemed to be self-sustaining, I stopped writing publicly about Weird Twitter. But I have been collecting data, as well as reading the research conducted by others. For the ethnographic results of this weird experiment, see part 2.

Tags: , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , ,

6 Responses to “Why Weird Twitter”

  1. fenton turner
    July 1, 2013 at 9:00 pm #



  1. Virtual identity edition: Subversive attention | Ethnography Matters - July 8, 2013

    […] Why Weird Twitter by Sebastian Benthall […]

  2. Complications in Scholarly Hypertext | Digifesto - July 12, 2013

    […] an opaque and problematic academic review process, I worked with editor Rachelle Annechino to craft a piece about Weird Twitter that was appropriate for the edition and […]

  3. On starting over, a year later. | emily grim - August 11, 2013

    […] you ever wanted to read a hyper-academic analysis of Weird Twitter? Here you go, don’t say I didn’t warn […]

  4. reflective data science: technical, practical and emancipatory interests? | Digifesto - September 4, 2013

    […] scientific techniques might be emancipatory (I was trying to accomplish something like this with Why Weird Twitter, for what its worth), but that was a qualitative study, not a data scientific […]

  5. The Anthropology of Weird Twitter | What can you do with anthropology? - May 1, 2014

    […] about how society works as well as teach important lessons about human behaviors. The article “Why Weird Twitter” by Sebastian Benthall takes a look at a particular internet community he defined Weird Twitter. […]

Leave a Reply