Elon Musk, Twitter, and the spam bot problem

Note: This was originally published as the daily newsletter for the Columbia Journalism Review, where I am the chief digital writer.

Elon Musk’s bid to acquire Twitter for $44 billion is only a month old, but it has already had more high-speed twists and turns than any Coney Island rollercoaster. After Musk filed notice of his offer with the Securities and Exchange Commission on April 13, Twitter’s board of directors implemented a “poison pill” defense, which would have flooded the market with cheap stock if Musk went ahead with his bid. Only a few days later, Twitter accepted his offer, in part because it was well above the stock’s recent trading price. This triggered a wave of speculation about what Musk planned to do with the service; among other things, he said that he would make the service’s recommendation algorithm public, and confirmed last week that he would reverse the permanent ban on Donald Trump’s Twitter account.

Then late last week (on Friday the 13th, no less) came a series of tweets in which Musk declared that his offer for Twitter was “on hold,” until he could verify the company’s recent statement that spam bots and other fake accounts make up less than five percent of Twitter’s total user base. At a technology conference in Miami on Monday, Musk expanded on this concern, saying he believed that the true number of spam or fake accounts could be 20 percent of Twitter’s total user base or higher, although he didn’t provide any evidence to support his estimate. Musk also said a deal for Twitter at a lower price “wouldn’t be out of the question” (Twitter’s share price is currently in the $36 range, more than 30 percent below where it was after Musk filed his offer.) The company responded that it plans to “enforce the merger agreement.”

Some observers believe Musk’s concern about the percentage of fake accounts is a ruse to either back out of the takeover deal, or at least negotiate a lower price. Matt Levine, an opinion columnist for Bloomberg, wrote recently that he doesn’t believe Musk really cares about spam bots. “I think it is important to be clear here that Musk is lying,” he said. Musk “has produced no evidence at all that Twitter’s estimates are wrong, and certainly not that they are materially wrong or made in bad faith,” Levine wrote. He added that the only way Musk could get out of the deal would be to prove that such a mistake would have a “material adverse effect” on the business, which he called “vanishingly unlikely” (although Musk did question whether advertisers are getting what they paid for, which he said was “fundamental to the financial health of Twitter”.)

On May 16, Parag Agrawal, Twitter’s CEO, posted a Twitter thread that appeared to be in response to Musk’s concerns, saying: “Let’s talk about spam. And let’s do so with the benefit of data, facts, and context.” Agrawal noted that spam “isn’t just ‘binary’ (human / not human)” and that the most advanced of these types of accounts use a combination of human beings and fully automated or ‘bot’ accounts, sometimes known as cyborgs. “They are sophisticated and hard to catch,” Agrawal wrote. “Many accounts which look fake superficially are actually real people. And some of the spam accounts which are actually the most dangerous [can] look totally legitimate.” Agrawal added that the estimate of fake accounts is based on random human reviews of accounts. In response, Musk posted a tweet consisting only of a poop emoji.

Newsweek wrote that close to 50 percent of Joe Biden’s followers appear to be fake, based on an assessment by a spam monitoring company called SparkToro. The same company reportedly found that more than 70 percent of Elon Musk’s followers appear to be fake, based on a number of factors such as their user photos and activity levels. Agrawal, however, said in his thread about fake accounts that only Twitter can be sure who is fake and who isn’t, because only it has access to private data about the account, such as IP address, phone number, location, etc. “FirstnameBunchOfNumbers with no profile pic and odd tweets might seem like a bot or spam to you,” said Agrawal, “but behind the scenes we often see multiple indicators that it’s a real person.”

One of the issues with the topic of spam, bots, and other fake or “inauthentic” accounts (as Facebook likes to call them), is that these terms are often used interchangeably, and inaccurately. Since the 2016 election, there has been a lot of commentary about the effect that “Russian bots” allegedly had on the outcome, but the term “bot” is typically used to mean an automated account that tweets without human intervention. Many of the Russian accounts referred to were run by human beings, although they may have been using fake identities, and engaging in inauthentic behavior. Many were likely part of organized “troll farms” like the one Adrian Chen described for the New York Times in 2015, known as the Internet Research Agency, run by an associate of Vladimir Putin.

The Internet Research Agency and its ilk often use automated accounts as well, to retweet or repost messages they are trying to distribute. But those are also the easiest to find, since their behavior is obviously fake. Both Twitter and Facebook take down millions of these kinds of accounts every month. What is harder, as Twitter’s CEO suggested in his tweet thread, is determining which accounts are fake or automated, and which are run by human beings who just happen to act like spammers or bots—whether because they are obsessed with a certain political viewpoint, or because they are pretending to be for other reasons, such as generating revenue. Whatever his motive for raising the issue, the reality is that Musk’s inquiry doesn’t have a simple answer.

Here’s more on Twitter and bots:

Bot people: A recent episode of the Reply All podcast looked at rumors that Christina Pushaw—press secretary for Ron DeSantis, the governor of Florida—has an army of Twitter bots at her disposal that she uses to target people online. But in the end, the podcast came to the conclusion that many of the most vocal accounts who retweet her criticism or attack her opponents are run by human beings, some of whom just want to be part of a larger movement. “Even the folks at Bot-o-meter admit that bot detection is not an exact science,” the podcast reported. “It’s why the software doesn’t actually tell you outright if an account is a bot or not. Instead, it just gives you a probability.”

Depp fans: The trial involving actors Johnny Depp and Amber Heard has become a hotbed of bot-like activity, to the point where both sides of the case have accused the other of employing “fake fan armies” to attack their opponent online. But are the accounts that tweet the most about the case actually bots? “The reality is far more disturbing,” writes Rolling Stone magazine; they appear to be mostly real people, some of whom post extremely offensive commentary. Dan Brahmy, the CEO of Cyabra, a startup that tracks bot activity, told Rolling Stone that “the majority of the profiles that are on Johnny Depp’s side and support him are real people, almost 95 percent of them.”

Bad algorithms: In a recent tweet, Musk suggested that users of Twitter should switch their feeds to a chronological view, rather than one that shows tweets chosen by the company’s algorithms. “I’m not suggesting malice in the algorithm, but rather that it’s trying to guess what you might want to read and, in doing so, inadvertently manipulate/amplify your viewpoints without you realizing this is happening,” Musk said. He also recently posted that he has switched his allegiance to the Republican party. “In the past I voted Democrat, because they were (mostly) the kindness party,” he wrote. “But they have become the party of division & hate.” Musk suggested he would soon be the victim of a “dirty tricks campaign” as a result of his political preferences.

Just business: Despite the rumors and confusion that have been flying around the deal over the past few weeks, Bloomberg reports that behind the scenes “it’s business as usual, as advisers on both sides plug away at the day-to-day work of closing a megadeal.” On Tuesday, a document was filed with securities regulators detailing how the offer came together, which Bloomberg said was “the result of weeks of coordinated work by both Musk and Twitter’s teams, according to people familiar with the matter.” Musk reportedly signed off on the final version—including the deal price of $54.20 for each share of Twitter’s stock—before the document was filed, Bloomberg reported.

Other notable stories:

Only three weeks after it was announced, the Disinformation Governance Board has been “paused,” according to employees at the Department of Homeland Security who spoke to the Washington Post. The paper said the decision came at the end of a “back-and-forth week of decisions that changed during the course of reporting of this story.” On Monday, the DHS decided to shut down the board, which had been attacked by right-wing commentators in what seemed to be a coordinated effort. Nina Jankowicz, the executive director, wrote a letter of resignation, but on Tuesday, it appeared the board might not shut down. Then Wednesday, it was put on pause, and Jankowicz officially resigned.

Residents of Illinois are getting checks worth $397 each from Meta, the parent company of Facebook, as part of a settlement that the company entered into to settle a class-action lawsuit over facial recognition, NBC News reported. The lawsuit alleged that the company violated the rights of  Illinois residents by collecting and storing digital scans of their faces. This is a breach of the state’s Biometric Information Privacy Act, which NBC said is “an unusual law passed in 2008 that allows consumers in the state to sue companies for privacy violations involving fingerprints, retina scans, facial geometry and similar data.”

Alex Goldman and Emmanuel Dzotsi, co-hosts of the Reply All podcast, will be leaving the show next month, according to an internal memo obtained by The Verge. Executives at Gimlet, the company that produces the podcast, emailed staffers Wednesday morning, saying, “Alex and Emmanuel have both made the decision to explore opportunities outside of Gimlet, and we’ll have more info to share on the future of Reply All soon.” The Verge reported that Dzotsi confirmed he and Goldman are leaving the show.

Chris Wallace will host a Sunday night interview show for CNN starting this fall, the network said Wednesday, according to Associated Press. The show, which will be called “Who’s Talking to Chris Wallace,” was the program Wallace was originally supposed to do on CNN+, the streaming service that was launched earlier this year and then quickly shut down, after just a month, by the network’s new owners, Warner Bros. Discovery.

The Chinese government routinely deletes posts, suspends accounts, blocks keywords, and even arrests those who post forbidden content on social media. But now they are trying a new trick, according to the New York Times: they are displaying the locations of those who post on social media. “Authorities say the location tags, which are displayed automatically, will help unearth overseas disinformation campaigns intended to destabilize China,” the Times reports. “In practice, they have offered new fuel for pitched online battles that increasingly link Chinese citizens’ locations with their national loyalty.”

In Muck Rack’s report on the State of Journalism for 2022, the site says that 77 percent of journalists value Twitter more than they do any other social media platform, and almost 40 percent of those surveyed said they plan on spending more time on Twitter this year than they did last year. Journalists ranked Twitter as one of their top destinations for finding news, second only to online newspapers and magazines.

Russia said it is closing the Moscow bureau of the Canadian Broadcasting Corporation, Canada’s publicly-funded news network, and will strip the CBC’s journalists of their visas and accreditation, the broadcaster reported. The network said the move appeared to be in retaliation for Canada banning the Russian state TV station, RT (formerly Russia Today.) Canada’s telecommunications regulator removed the Russian station from its list of authorized distributors in March.

Recurrent Ventures, which owns magazines such as Popular Science and Outdoor Life, said Wednesday it has closed a $300 million round of new financing. The company said it has raised a total of $400 million since it was created in 2018 by Andrew Perlman and Matt Sechrest, former partners in a private equity fund. In addition to Popular Science and Outdoor Life, the company owns the home improvement site BobVila.com, the military magazine Task & Purpose, and the outdoor magazine Field & Stream.

Leave a Reply

Your email address will not be published. Required fields are marked *