A stream of recent sneaky tweets and social media posts tell people they can “vote from home” by simply sending a text message, a devious tactic to suppress votes.
The U.S. election is not “American Idol,” and voters cannot – repeat CANNOT! – cast ballots by texting from their cellphones. Twitter says it is taking the tweets down.
The last-ditch appeals underscore a broader issue of concern in an especially contentious political year: the increasing usage of robotic networks, or botnets, to flood the internet in an attempt to influence the election, squelch public debate, spread lies and manipulate voters.
Researchers say highly automated social media engines are working for both major presidential candidates, posing as real people when they are actually just machines. Often, it is difficult to discern who is behind the most diabolical of the social media messages.
“It might not be candidates or campaigns launching the bots,” said Samuel Woolley, the director of research at the Computational Propaganda project at Oxford University in England.
In recent days, a series of professional-looking tweets and Facebook and Instagram posts have encouraged supporters of Hillary Clinton to “Save time, avoid the line.” One of the posts has a photo of a confident-looking Clinton with the following: “Vote from home. Text ‘Hillary’ to 59925 and we’ll make history together.” Another has a similar content with a photo of an African-American woman.
The intent of the messages is to keep targeted groups – in this case, Clinton supporters – fooled into sending pointless text messages and not bothering going to the voting precinct, a technique designed to stop them from casting ballots.
Researchers say bot campaigns are versatile and ever more sophisticated.
“It’s an arms race,” said Filippo Menczer, a computer scientist at Indiana University’s School of Informatics and Computing. “There is an incentive for people who want to affect public opinion to generate smarter bots.”
Targeting people through tweets and social media posts is the modern iteration of political junk mail. Such campaigns are worth it even if you sway only a few people.
“Even if the yield is very low, the cost is so low that it makes it worth it,” Menczer said.
The lies can come quick and dirty.
“It’s a lot easier to trick you in 140 characters than in a face-to-face conversation,” Menczer added.
For those looking to harness Twitter battalions for political warfare, armies of dormant accounts can be conscripted with a little cash.
“You can buy a thousand accounts for $500 or $600 from a company in Singapore,” said Philip N. Howard, a political scientist and information expert at the Oxford Internet Institute.
Twitter says it has 313 million active monthly users, and it works hard to shut down active bots. But tens of millions of dormant accounts also exist.
“Some 20 to 40 percent of accounts are dormant, have no photos and tweet rarely on a wild range of topics. These accounts, we think, are bots,” Howard said. “They’ll have thousands of followers and not follow anybody. These bots often follow each other.”
Howard said “bot armies” behind both major presidential campaigns “have spread misinformation” and that “it’s time for some kind of public policy oversight.”
“It’s not good for democracy. It’s not good to have a tool that is so easily used to spread panic and misinformation,” Howard said.
Woolley, Howard and a third researcher conducted a study of tweets at the time of the three major presidential debates in late September and early October.
“We gathered about 9 million tweets for each debate,” Woolley said. “During those debates, one-third of all the traffic tweeting for Donald Trump was automated, driven by bots, and about one-fifth of all the traffic tweeting for Clinton was driven by bots.
“So you’re getting to the neighborhood of millions of tweets being driven by automated accounts. You could call them fake accounts.”
By the third debate Oct. 19, the researchers found seven times more bots tweeting for Trump than for Clinton, he said.
The research team used a variety of criteria to determine whether an account was automated.
“If an account tweets several times a minute, if it tweets thousands of times a day and if the time stamp for the account is a month old or two months old and it’s already tweeted several thousand times, it’s a good indication that it’s an automated outcome. It’s outperforming what a human could do,” Woolley said.
Some automated accounts – known as cyborg accounts – occupy a gray area.
“There are oftentimes a lot of humans sitting in a room managing thousands of bot accounts. When a bot gets into a conversation, it’s pretty easy for a human to step in and do the talking,” Woolley said.
Botnets have been deployed in a variety of countries to squelch dissent.
“We’ve seen examples in other countries – in Russia, Iran and in Mexico – of bots used to destroy social movements,” Menczer said. “They would impede conversations. . . . All of a sudden, you would see hundreds of thousands of junk tweets flooding your feed.”
In Mexico’s case, a hashtag became popular in late 2014 following the disappearance and presumed murder of 43 students at a rural teachers’ college. Picking up on the attorney general’s utterance to end a news conference, angry social media users employed the hashtag #yamecanse, or “I’ve had enough.” But the protest was partially hushed when bots used the hashtag to post innocuous soccer scores or other useless content, halting the conversation among users.