Radicalism and Internet Culture
Does online performance logically devolve into rage and violence?
Somewhat apropos my post about a seemingly normal conservative blogger turned ‘Stop the Steal’ agitator, New York Times media columnist Ben Smith shares a tale about a former Buzzfeed employee who stormed the Capitol. The particulars aren’t that interesting to me but the takeaway certainly is:
His real name was Anthime Joseph Gionet, though he preferred others. His value to BuzzFeed was clear: He’d do anything for the Vine, the short video platform that had a brief cultural moment before being crushed by Instagram and Snapchat in 2017.
Once, he poured a gallon of milk on his face and a clip of it drew millions of views, back when mostly harmless stunts amused millions of American viewers on the platform.
[…]
Mr. Gionet was standing inside the trashed office of Senator Jeff Merkley of Oregon, streaming from one of the few platforms yet to ban him, alongside other Trump loyalists who played with the telephone receiver and draped themselves over the furniture. It seemed an apt conclusion to a recent career arc that some might see as trolling or internet pranks, but is probably best described as performative violence.
[…]
When he left BuzzFeed later that year to work as the “tour manager” for Milo Yiannopoulos, a darling of the racist and anti-Semitic “alt-right,” colleagues were momentarily shocked. Then, they scrolled through Mr. Gionet’s Twitter account, where his increasingly vile statements were getting him retweets from far-right figures, and realized that they shouldn’t have been.
Still, it’s not clear what Mr. Gionet actually believes, if anything. And really, I’m not sure I care.
This isn’t a sympathetic profile of a young man gone wrong. I can’t muster much pity for a guy who, before he was attacking his Capitol, spent his time shooting some kind of bottled irritant (he called it “content spray”) into the eyes of innocent people for YouTube views and shouting at store clerks who asked him to wear a mask.
To me, this story is about something different, a sort of social media power that we helped sharpen at BuzzFeed that can exert an almost irresistible gravitational pull.
If you haven’t had the experience of posting something on social media that goes truly viral, you may not understand its profound emotional attraction. You’re suddenly the center of a digital universe, getting more attention from more people than you ever have. The rush of affirmation can be giddy, and addictive. And if you have little else to hold on to, you can lose yourself to it.
[…]
Even as we sought to make our work spread at BuzzFeed, we faced constraints — by truth in our news division, by hewing to a broadly positive set of values on our entertainment side. But Mr. Gionet ultimately broke loose of those boundaries, seeming to follow the signals he found on social media without any scruple. The only through line was his desire to build an audience. He was boosting Bernie Sanders before he was chanting anti-Semitic slogans in Charlottesville, Va., then temporarily recanting those extreme views and later committing violent crimes to get views on YouTube. He built an audience among coronavirus deniers and then, when he apparently contracted the disease, posted the screenshot of his own positive test to Instagram with a tearful emoji. A few weeks later, he joined the pro-Trump uprising in the Capitol.
“His politics have been guided by platform metrics,” reflected Andrew Gauthier, who was a top video producer at BuzzFeed and later worked for Joseph R. Biden Jr.’s presidential campaign.”You always think that evil is going to come from movie villain evil, and then you’re like — oh no, evil can just start with bad jokes and nihilistic behavior that is fueled by positive reinforcement on various platforms.”
In the early days of blogging, there was certainly a thrill to getting linked by a big-time blogger, which would send thousands of views in a matters of hours to a site that might only get dozens a day. And a lot of us spent a lot of time generating clickbait to keep those highs coming.
Whether because of my academic and military training, concern for reputation, or simple wiring of personality, I eschewed inflammatory writing about the subjects I cared about. But I wasn’t above gaming the system to bring in fresh hits from other bloggers, climb the ranks of the Ecosystem, generate clicks from search engines, and the like. I was posting caption contests, linkfests, traffic jams, Blog Chick Pics, and other gimmicky crap before Buzzfeed was even a thing. For a variety of reasons, I gave up on those things a very long time ago but I can certainly see how someone wired a little differently—and with less going on in their personal and professional lives—could get sucked into the vortex.
And, even aside from all that, I recognized very early in my blogging days—and I’ll hit 18 years at the end of this month—that all the incentives were toward extreme views. While I was on several of the “Conservative Bloggers Liberals Should Read” lists, the numbers of liberals interested in reading my thoughts was a tiny fraction of rabid conservatives I could have garnered with my early-mover advantage. It just wasn’t a game I was wired to play.
His story leaves me wondering what share of blame those of us who pioneered the use of social media to deliver information deserve at this moment. Did we, along with the creators of those platforms, help open Pandora’s box?
I didn’t work directly with Mr. Gionet. But in 2012, I did hire a writer named Benny Johnson who was cultivating a voice that blended social media savvy and right-wing politics. I thought, wrongly, of his politics at the time as just conservative. And I imagined him thriving, as conservative writers have done for generations in mainstream newsrooms, where they shared their colleagues’ interest in finding shared facts.
But Smith goes much further:
I was slow to realize that his interests weren’t journalistic, or even ideological, as much as they were aesthetic, thrilled by the imagery of raw power. In the tradition of authoritarian propagandists, he was awed by neoclassical buildings, guns and, later, Donald Trump’s crowds. And, after we fired him for plagiarism in 2014, he went on to lead the content arm of Mr. Trump’s youth wing, Turning Point USA, and host a show on Newsmax. Last week, he was cheerleading attempts to overturn the election (though he pulled back when the violence began and later blamed leftists for it). He’s also selling his skills in the “viral political storytelling” that we worked together on at BuzzFeed to a generation of new right-wing figures like Representative Lauren Boebert, who has won attention for vowing to bring her handgun to work in Congress. (Neither Mr. Gionet nor Mr. Johnson responded to email inquiries.)
While we were refining the new practice of social media at BuzzFeed, we were slow to realize that the far right was watching closely and eventually imitating us. Jonah Peretti, who founded The Huffington Post as well as BuzzFeed, was surprised when Steve Bannon, who ran Breitbart, recalled to a writer that he’d borrowed elements of his strategy from Mr. Peretti in the run-up to the 2016 election. Mr. Bannon told me before that election, in an interview in Trump Tower, that he was surprised we hadn’t turned BuzzFeed to pure Bernie Sanders boosterism, as Breitbart did for Donald Trump. He noted, probably correctly, that the traffic for a pro-Sanders propaganda outlet would have greatly exceeded what we got for fair coverage of the Democratic primary.
“Some of the innovative things we did early on, in understanding social media and digital media, have been taken up by alt-right groups, racist groups, MAGA groups,” my old boss, Mr. Peretti, told me in an interview last week. But Mr. Peretti, an eternal optimist, noted that some of the same social mechanisms that Mr. Gionet exploited were also crucial to the sweeping progressive social movements of the last few years, from Black Lives Matter to #MeToo. “The story’s not done and there’s an opportunity to fight for a good internet,” he said.
I’ll need more evidence before concluding that violence is the natural evolution of Internet performance but it’s more than a little plausible. If nothing else, keeping a white hot interest in politics outside of campaign season is really helped if the other side is seen as A Danger to the Republic and Objectively Evil. And, as we’ve seen, that makes losing an election akin to losing a war.
I’m already hearing what seem to be two competing explanations of what happened in Washington last week: that the overwhelmingly white, sometimes overtly racist, mob embodied old, deep unexpurgated American evil; or that social media reshaped some Americans’ blank slate identities into something radical.
But Mr. Gionet’s story shows how those explanations don’t really conflict. A man his colleagues saw as empty and driftless turned his identity into a kind of a mirror of that old American evil, and has become what many Americans told him they wanted him to be.
At one point in Mr. Gionet’s livestream during the siege of the Capitol, an unseen voice off camera warns that President Trump “would be very upset” with the antics of the rioters.
“No, he’ll be happy,” Mr. Gionet responded. “We’re fighting for Trump.”
Reportedly, Trump was happy before he wasn’t. I suspect Gionet and his ilk aren’t “classy” enough.
We don’t normally blame the Lumière brothers for Triumph des Willens or Birth of a Nation.
If Trump was unhappy, it was because their antics got him in trouble and banned from Twitter, etc. He always blames someone else for his issues. Of course, he’d complain that they look trashy. If the protest had driven lots of news stories that weren’t about violent trespass and police officers getting killed, he would still be happy with them.
Social media forms a Skinner Box, which appears to influence what people do offline as well as online.
@drj: Blame? People are still responsible for what they do, even if they cry “I didn’t mean to!”. I didn’t mean to drive over 70 mph on that Texas Highway a couple years ago, but I still got a ticket. (Seriously, my speedometer said 70).
This is more like we are noting that there’s a big downhill slope that really, really increases the number of people speeding. It doesn’t force them to speed, they could use the brakes, but many don’t.
I’m kind of afraid to ask… but what were Blog Chick Pics?
Does the medium affect the message.
Whenever this comes up, my mind always goes to this passage from Walter Benjamin’s “Work of Art in the Age of Mechanical Reproduction.” After talking a bit about the revolutionary aspect of film is the ability for it to be used to capture and circulate the images (and with sound the opinions) of the general public, Benjamin writes that:
I think there really is something to this (or rather how we are wired as a herd-based species and the impact of communication mediums that allow “average” folks to speak to the broader population–see talk radio as a prime example of this).
@wr: Literally links to (perfectly tame) photos of female bloggers hosted on their blogs. Inspired by a glamour shot of Megan McArdle getting a gushing response on her own site. Pretty wholesome in the grand scheme of things and it actually drove community and visits to their sites but it’s not the kind of thing I’d do now.
Would that be this Benny Johnson?
@MarkedMan: Yes. Both mentioned in parts of the article I didn’t quote.
Along these lines, the online journalism collective Bellingcat release a powerful piece last week on the last year or so of Ashli Babbitt’s life as it was reflected in social media, leading up (content warning) to her final moments during the siege.
https://www.bellingcat.com/news/2021/01/08/the-journey-of-ashli-babbitt/
The WaPo would later build on this for their own article.
I think we should think of political clickbait in the same way we think of Instagram lifestyle influencers. It’s not violence that’s the natural outcome, but a sort of picture-perfect existence. A person who wants their Instagram feed to be a constant stream of pure yoga poses and distressed Americana and endless displays of natural beauty is not any different than a person who wants to win every argument and have the most facts and logic.
The difference is that most people are aware that a spotlessly beautiful life is not only impossible but not worth having. Being judged as fat or out of style or basic is not the end of the world. People who wear normal clothing aren’t going around furious about how the editors of W think of them. But in politics there’s no chill factor, no sense that being perfectly consistent is way overrated, and zero confidence in one’s own mind.
@mattbernius:
What struck me from that article:
Babbitt’s descent into madness started with standard Fox News/talk radio fare.
I don’t think that a bunch Twitterati with numbers in their user names could be very successful without a more mainstream ecosystem to build on/grow out of.
Somebody with a bit more authority than “Patriot4Evah1234” has to validate their bullshit, or at least create the conditions in which apparent nonsense can be readily believed.
Not at all. Even Zuckerberg and the Google founders didn’t really understand what their platforms would do at the beginning. It seems they still don’t and are constantly playing catch-up.
I honestly miss the heydeys of blogging, I think it really was a golden age.
Social media platforms were, perhaps unintentionally, designed to activate deep and primal parts of our brain. The need for social acceptance and social status is a very powerful motivator for most people. I think that’s why so many spend their time chasing likes, retweets, etc. You also see it when someone disses someone else for things like a “low follower count.” Plus, a lot of it is addicting for many people. Add in political radicalism, the atomization of society, the focus on materialism as the most important moral structure, and it is a toxic brew.
I think social media also makes us lazy, or activates our lazy gene. So many people I know thinking that posting opinions on social media is “doing something.” Friends post long rants on Facebook about how everyone who doesn’t believe in X should unfriend them, but it turns out that ranting on Facebook is the extent of their outrage. They don’t give time or money to actually oppose X, don’t contact their representatives to give their opinions on X, etc. I’ve always found this to be puzzling behavior. So much of social media appears to me to be entirely virtue and status signaling.
I wish I had an answer for how to, collectively, make this healthier both for individuals and for society, but I don’t. Social media is driving the national conversation and it is dominated by a small minority of the population which seems to be unrepresentative of the whole.
Being judged as fat or out of style or basic is not the end of the world.
Let me amend that by saying that it’s considered wise to rise above that type of judgement. Middle-aged women who wear Ann Taylor aren’t being directed to fashion websites where assistant editors who wear Commes des Garcons are making fun of them. Whereas political clickbait is like a search for somebody out there anywhere who wants to call you a mook. You have people taking an op-ed by a trans person or a college kid’s twitter feed and blasting it out to middle-aged people in Wisconsin, and it’s considered completely reasonable to be outraged. There’s no wisdom at all.
@Andy:
Agreed. This is pretty much the story of every new/disruptive technology. Predictions on what it might be used for at the time of introduction often don’t mate the trajectory said technology takes*–in no small part because users are unpredictable and often find new and unexpected applications for said technology. There’s always a reflexive feedback loop that forms there.
It’s just that it often gets calcified over time as a given technology matures.
* – To get McLuhan for a sec, it’s because we understand most technologies with the metaphors of the past (as he put it we walk backwards into the future).
@Andy:
Yup, we’re here, when we could be doing something productive. 🙂
@Andy:
@mattbernius:
I feel it safe to say the next new, revolutionary technology that lives up to its initial promises and expectations will be the first.
Ok, that’s not true. But many such developments came along with bright visions of future improvement, and then fell not just far short of such goals, but in another realm altogether. I recall hearing how TV would revolutionize education (and I was born well after TV sets were as common in houses as floors and ceilings), then it was the VCR. Shopping malls were intended as places for people to socialize.
It’s far safer to say any new, revolutionary technology will evolve along the path of maximum profit, and even then predictions on how such profit will be made are bound to be wrong, incomplete, or merely a starting point.
What’s troubling is that no matter hos harmful a new technology proves to be, what determines its permanence is whether it makes money or not. Not the same, but didn’t we see that with prescription opiates?
@Andy:
100% and they’ve been optimized for that via algorithms and other tools. It’s pretty easy to optimize for if you consider any interaction and engagement with a platform good. Unfortunately, what we have seen is that the content that loads the highest about of affect, and therefore leads to higher degrees of engagement, tended to be content you disagree with.
The same is true of a lot of mobile “freemium” games (which like SM are optimized to trigger dopamine hits), just in slightly different forms (maximizing for surprise in those cases).
This is actually an ongoing debate within design communities about the need to actually take our wiring into consideration and build ethical friction into interfaces/experiences/services to try and lessen the impact of these tools.
The fault lies not in our media but in our brains. Once upon a time we had gatekeepers who obviated much of the need for epistemological rigor. Walter Cronkite was not out to lie to us. The New York Times was not out to lie to us. And then came Fox News, which was out to lie to us. The gatekeepers – never perfect, of course – became in some cases openly corrupt, dishonest, even malicious.
Social media just accelerated the work Murdoch, Limbaugh, Jerry Falwell and their ilk began. From generally reliable gatekeepers, to a mix of reliable and dishonest gatekeepers, to no gatekeepers. Fine for many people. Fine for people who had developed their own capacity to judge truth and falsehood, and were able to separate that from their own needs and desires. So, what, 5% of the population?
With gatekeepers gone or tarnished the 95% of the population that had never spent ten seconds considering the how of determining what is true was free to believe not what was true, not what was supported by evidence, but whatever the hell made them happy. And then they could repeat the happy lies and make more people happy with happy lies.
This affected the Left, too. We have our own anti-vaxxers. We have our own conspiracy nuts. And yes, we regurgitate and amplify any message we believe makes us look clever, or morally superior, no matter how idiotic.
But the Left in this country cannot match the unhinged idiocy of a Right wing already groomed by Fox News, talk radio and evangelical Christianity. The Left is often stupid, usually smug and obnoxious, but we remain in contact with reality because we held more firmly to reliable gatekeepers – academia, mainstream media, books. In part this is because the Left is, on the whole, simply more intelligent. But it’s also because we won the culture wars and the Right lost, so reality itself became dangerous to their beliefs while remaining fairly pleasant for us.
Winners like reality, losers don’t. That’s not hard to understand. The guy winning is not the guy yelling that the deck is stacked.
In any case, yes, Gutenberg helped to launch the Renaissance and accelerated the Reformation and all the fun religious slaughter that ensued, but that’s on Gutenberg. He didn’t make people stupid.
@mattbernius:
It’s pretty easy to optimize for if you consider any interaction and engagement with a platform good.
Exactly how do they know it’s being optimized? Genuinely curious here. We’re treating social media as if it has cracked the code of our brains. I’ve been blessed with good mental health, but I’ve known enough people who have struggled with mental illness and medication. It’s not like the doctors have figured out to optimize mental health with meds. There’s often a huge struggle to make it work, and sometimes it doesn’t.
By a fluke of history, the riotous mob that spawned the Boston Massacre are labeled freedom fighters while John Wilkes Booth was villified when he knew he was the American Brutus. As with most senseless violent acts through the ages the final verdict is determined by the victors. In both cases I believe these were willing dupes; oh how history repeats itself, even if the delivery platform is different.
The natural outcome, if not evolution, of the internet isn’t violence.
It’s sorting. People sorting themselves into echo chambers where one need never entertain a contrary thought, only anecdotally be disagreed with, and above all never have to be troubled with the presence of “other” (be that whatever outside the group demographics it may be).
The internet, at its most basic I think, makes it a great deal easier to ignore, then dehumanize one’s opposition. The violence follows from that as night follows day.
@HarvardLaw92: I would argue that we always have had the “sorted” media (different newspapers addressing different markets)–but that it used to be that any newspaper that wanted to survive had to have at least some connection with reality. Just as the readers of the WSJ are going to get extremely indignant if at least the business section of the journal doesn’t produce data and reports that they can rely upon. So there might be some cranky ideas put forth here and there, but on the whole, the stuff had to be….sensible.
The problem with the advent of the internet is a) no gatekeepers anymore and b) it’s now possible for people looking for craziness to get the stuff by the ton, neat and unadulterated, and piped directly into their veins.
I’m sure that this phenomenon has multiple complex causes, but I think Rush Limbaugh was a big impetus. In 1988, there was a conservative commentator on my local radio station during morning drive time who was serious, calm, and analytical. He was also local. He was replaced by Rush who was syndicated. I’m sure that Rush was a lot cheaper. At first Rush was amusing; I saw no depth, but he was frequently funny. He was no fan of Bush the older who was not liked by hardcore conservatives, but he became totally partisan by the 1992 election and turned into a 100% GOP advocate. I stopped listening then; there is no point to listening to an unending series of ads for Budweiser even if you like the beer and Spuds MacKenzie. Somehow Rush found that people have an unsatable hunger for 4th grade humor. Rush gave his audience the first rush of being in an inner circle of special insightful people, and they keep upping the dose to get that feeling.
@Modulo Myself:
Short answer is I work in the tech and research space. And so the issue of algorithmic optimization has been something that’s been studied for years (both within and outside) of these orgs. This also includes cases where the algorithms have optimized for the wrong goals. Or rather the definition of something like “engagement” didn’t necessarily take into consideration additional vectors like positive versus negative interaction. For quite a while it would just push content that you were more likely to react to (it’s why, for a period in the late teens, there was a high probability you were more likely to see connections content that you would find controversial versus stuff that you would agree with). Basically machine learning led it to see pushing controversial content as better than agreeable content (here’s a parallel example of what happens with machine learning and dubious training data/parameters – https://www.theverge.com/2016/3/24/11297050/tay-microsoft-chatbot-racist).
Facebook had to rework the algorithm in part because of that.
This article from Quartz touches on some of the issues with affective loading:
https://qz.com/1039910/how-facebooks-news-feed-algorithm-sells-our-fear-and-outrage-for-profit/
This is a prime example of what Andy is talking about re: not understanding the ramifications of a new technology until after it’s already had an observable effect.
@grumpy realist:
Again, this is a bit of an ahistorical argument that came out of the golden age of mass media (circa 1945 – 1990ish). Prior to that (and even through that period) papers were often hyper-partisan and sustained through circulation as much as advertising.
What happened during that period was consolidation and audience growth led to a windowing of the smaller fringe papers (though they still existed) and a move towards a professionalization of journalism. I’m a broken record on this, but what most of us grew up with was a historical aboration in terms of journalism and not the norm.
@mattbernius:
The problem with that, is that compared to today’s media we all know there is a better way. That makes today’s situation harder to take, not easier, even if it is a return to the mean.
We can say the exact same thing regarding income inequality. It decreased for much of the 20th century, then it widened again. Huge disparities in income are the historical mean for most human civilizations. This makes few people content with their loss of purchasing power compared to their parents and grandparents.
@mattbernius:
When you say that newspapers were at one point hyper-partisan, you are saying they generated stories that we now call yellow journalism, am I somewhat correct (I ask if I got it kind of right before I turn to Google to look up yellow Journalism)?
@grumpy realist:
I’m always a bit wary of the whole “This is like no moment that ever came before it” thing. The militia movement really began during Reagan’s time, and Timothy McVeigh’s bombing occurred in 1995. The World Wide Web wasn’t even made available to the public until 1991 and by 1995 was only starting to become a thing after the Mosaic browser came out. In 1995 most people’s awareness of the Web, such as it was, came from skeptical news stories wondering if it actually had a future. But despite this lack of social media there were still tens of thousands of fat slobs carrying AK-47 clones around, CosPlaying as imagined heroes and bragging about what they and their guns would do if those “urbans” came around. Basically, the same group of losers we have now. Sure, today’s thugs use social media to organize, but somehow lynch mobs were still very successful for many, many millennia before smart phones arrived.
@Andy:
Facebook has been used to organize and incite genocide.
“Even Zuckerberg” suggests that he was trying to avoid these bad outcomes, but the truth is he clearly didn’t care. If he cared, he would have pulled the plug on Facebook in Myanmar when it became apparent what was happening.
If there is a hell, Zuckerberg will burn for eternity. Since there is likely not a hell, the best we can do is hope one of his mansions burns down with him inside it. He’s a filthy, sociopathic little turd blossom.
My kingdom for an edit button. My last should have read something like “the web was only starting to become a thing in 1995 when McVeigh was already deep deep into his mania. In fact it didn’t even make it into general geek awareness until 1993 when the Mosaic browser came out”
Even Larry Page and Sergei Brin, with their whole “organize all the world’s information” would regularly make exceptions for information that would hurt people. They delisted the entire “See if your neighbors have an arrest record” industry once they (belated) realized that it was just being used to blackmail people who didn’t want their arrest record following them.
(Contemplating starting to look for a new job, and a Facebook recruiter contacted me. No, I’m not going to work for your terrible company, not unless I am there to commit acts of sabotage and handing sensitive data over to reporters… and frankly, I don’t need that level of stress in my life…)
Basically machine learning led it to see pushing controversial content as better than agreeable content
This is basic tabloid journalism though. Machines didn’t need to figure this out. Humans got there a long time ago. What a platform like Facebook is ‘optimizing’ is the stream of delivery of content to a person who clicks on that content. They haven’t cracked a code that humans hadn’t already cracked. They just have tons of their own data about what’s clicked on AND humans spend an ungodly amount of time on the platform in its various guises so there’s a need for 24/7 content AND both delivery is easy. Also, tech people tend to be kind of clueless about human behavior, so what’s common knowledge is reframed as a sinister breakthrough accomplished only by social media.
@Modulo Myself: The big machine learning advances are in detecting the hateful content early so you can promote it earlier. Also, finding the gateway hateful content for different types of users.
You don’t get to lizard people in one step — it’s a long journey. But, with good machine learning, we can reduce the number of steps and turn your mildy racist grandmother into a gun toting loon who wants to kill her some lizard people in three months flat.
I mean, we can optimize your grandmother’s experience so she has a positive interaction with more content that generates ad impressions.
A forensic psychiatrist expounds on Trump and his cult.
But, with good machine learning, we can reduce the number of steps and turn your mildy racist grandmother into a gun toting loon who wants to kill her some lizard people in three months flat.
Even as hyperbole, I think that’s bullshit. Machine learning is an ambiguous with weird results. The idea that a bunch of code has mastered causal pathways is absurd. The most popular Pavement song on Spotify, for example, is a B-side from Brighten The Corners. It’s completely forgettable. When Stephen Malkmus walked into a cafe where it was playing he didn’t remember it. This is the work of Spotify’s algorithm. It’s funny, because Pavement fans like me are the type of people who deserve being stumped by the existence of this song and I kinda hope it is the Pavement song for the next generation. But the idea that machines are keepers of deep secrets about what works that we can’t figure out is bullshit.
Bottom line is that any grandmother who goes on a gun-toting rampage was repressing this for a very long time. It’s like these anti-trans idiots who believe the internet made their child trans overnight. They can’t face the fact that their child didn’t trust them and didn’t talk to them, so they invent a radicalization on Youtube rather than the sad truth.
@Modulo Myself: The Pavement song was popular on TikTok, presumably because of the ridiculous lyrics. Machine learning has nothing to do with it.
That’s good shit in a weird monotone sing song. (I have figured out why YouTube is recommending silly, fun songs to me every time I finish a video on physical therapy… It has decided I am a 15 year old with the knees and back of a 50 year old, and who wants TikTok songs…)
The question is why this song was never a hit before — it doesn’t feel like a standard hit song. But once a couple of cool kids on TikTok found it and used it for videos, of course it would be popular.
You can absolutely get someone to believe insane stuff by adding a few bits of crazy each step of the way, especially if they find a community around it. And then, convince them that community is under attack, and they will try to defend it. See most religions.
Now, what engages people online? Something they generally agree with that triggers an emotional response. Something 3% crazier than the person is. And when their engagement starts to wane, you need to ramp up the crazy a little because the person has incorporated the crazy. And it’s personalized. What works for one person won’t work for everyone, and that’s where the machines come in — to categorize content and test the person regularly with things a little more crazy in various directions, backing off if engagement drops.
@Modulo Myself: Alternately, reading the article more, Spotify did the same thing with “Harness Your Hopes” that Facebook does with “Your Jew Neighbors Are Secretly Muslim” (or whatever). It carefully experimented to find something that people would like even though they didn’t think they would like it through a mixture of brute force and machine learning.
You can’t argue with the results of the song. 15 year olds are now singing along to a song about murdering your parents and roasting them on a spit. They never set out to do that (apparently… if the dates in the article are right), and yet here we are.
Just as grandma never set out to worry about lizard people who took us off the gold standard.
@Modulo Myself: Thinking while heating up dinner, I’ll add this:
1. No, computers did not discover something about people. The radicalization is a side effect of optimizing for engagement, and it’s something very basic and predictable about people that ends up getting exploited.
2. Facebook and YouTube don’t, in all likelihood actually plan on getting your grandma to believe in lizard people. They do, however, adjust recommended content as grandma’s interest wanes, which creates a feedback loop. It can go in any direction, generally, and YouTube doesn’t care if you’re watching videos about Star Wars history, or how white people are being oppressed.
3. Natural language processing in common languages is good enough for categorizing content. It’s not just people who liked this black-box also liked this other black-box — they can peer inside the black box. They can absolutely categorize videos by content and find the medium-conservative anti-immigrant content that avoids racial slurs but speaks to those grievances. This is where a lot of the machine learning goes, and it’s pretty good.
4. Each recognized user is going to have a set of interest clusters (TikTok-videos with upbeat sounds, medium-conservative anti-immigrant without racial slurs, physical therapy about knees, Star Wars content about vehicle designs), and when engagement starts to lag, the system will experiment adding a small percentage of variant content (medium-conservative anti-immigrant with a few racial slurs, more conservative anti-immigrant, immigrant stories about medium-conservative immigrants who love Star Wars vehicles, something about a wall, how do you know if you need knee surgery, etc.)
5. The slightly more radical almost always performs better.
6. More ad impressions. And that’s where the company’s interests usually end.
Now, a clever person can absolutely tweak the adjustments in step 4. You might have a limit on how far you will demonize immigrants, or you might recognize that people with bad knees who searched for car reviews three years ago is going to be interested in red cars, or you might pull up a few things about the causes of immigration, or you can take your hands off the wheel and let them slink over to the lizard people… you can absolutely steer where the person will go next, just less optimally.
Step 4 can also be adjusted through machine learning to add or slow the person’s content drift, anticipate where they are going, and get them to somewhere different but in a similar direction (similar measured against the user — see all the folks who bounce between hard left and hard right, where their views seem unpredictable unless you are just looking for crazy town)
That step there, where the decisions are being made about what to present to the user next, to choose which user-generated content… I would take away Section 230 protection for that, as it has nothing to do with user content. But that’s another issue.
Anyway, you might still say that this only works if there is a seed already in the person that can be exploited. And you might be right, but show me a white person who doesn’t have racial resentment about losing their privilege.
It’s a function of the newness of the medium. When printing presses became common, pamphlets were capable of enraging and enlightening people and helped lead to the breakdown of the old social order.
When radio came on the scene, people just didn’t have the experience to filter what they heard. Famously, some citizens overreacted to a presentation of “War of the Worlds,” afraid of Martians with heat rays. Hitler tapped into the new potential, as did Churchill and Roosevelt for quite different purposes. They moved entire nations.
With television there was McCarthy, Kennedy-Nixon, and you just had to get the right deodorant. But people eventually became inured to the message.
Then came cable TV, blogs, and social media in turn. When a medium is new, most people don’t have the ability to separate the wheat from the chaff. Over time, most will develop better discernment. Then new media will come along to be used and abused.
@inhumans99:
Yellow journalism was an aspect of that (and Hearst stuff was honestly mild compared to some–some Northern Conferdate-sympathizing Newspapers during the Civil War were essential making up stories about how the South was really winning during the latter half of the war).
And in addition to that, the reason that many newspapers have “Democrat” or “Republican” in their names was literally because they were wings of local political parties (and not just the papers with those titles in the names). Those connections were with Newspapers more or less from the beginning in the US.
For more reading on that, I highly recommend “Perilous Times: Free Speech in Wartime” by Geoffery Stone. While its focus is specifically on Free Speech in Wartime, that topic is so bound up with the press he has to trace the history of journalism (starting with print) from its beginnings in the colonies.