A wise man once offered some sage advice:
Unfortunately, Twitter bots can’t heed this warning, because they’re programmed to do the opposite. And this campaign season, many, many bots have been programmed to tweet about the election. So many that researchers found that bots make up 19% of the Twitter discussion about this year’s presidential contest.
A new study published in the internet-focused journal First Monday analyzed activity on Twitter over a number of weeks between September 16 and October 21, tracking 23 hashtags and keywords, including #ImWithHer, #NeverHillary, debate, #jillstein, #garyjohnson, #trump, and #tcot.
Alessandro Bessi and Emilio Ferrara, the two University of Southern California researchers behind the study, arrived at a sample of “7,112 Clinton supporters (590 bots and 6,522 humans) and 17,202 Trump supporters (1,867 bots and 15,335 humans)” by following the top few hashtags users tweeted and whether those hashtags were generally positive or negative towards Clinton and Trump. They used BotOrNot, a program designed at Indiana University, to figure out which accounts were bots.
Then, they wrote, they used the activity of these users, human and bot, to estimate the behavior of the election-discussing Twitter population:
By extrapolating for the entire population, we estimate the presence of at least 400,000 bots, accounting for roughly 15 percent of the total Twitter population active in the U.S. Presidential election discussion, and responsible for about 3.8 million tweets, roughly 19 percent of the total volume.
This isn’t as surprising as it may seem. Bots are ubiquitous on Twitter, and have been for years. In 2014 the company said they made up 8.5% (23 million) of the platform’s total active users. Twitter bots can be bought en masse as followers, or made wistful, fun, and eerily human. Left unprotected, they can also be turned into neo-nazis.
Even more specifically, they’ve already dominated Twitter this election (or at the very least towards the end of the election). In October Philip Howard, an Oxford professor studying bots and political discourse, released a paper that found that a quarter of tweets about the first two presidential debates appeared to be bot-generated.
Bessi and Ferrara also discovered some interesting bits of information about the bots that supported various candidates. Clinton supporters who were human tended to be more positive about their candidate than her bot supporters. Trump supporters on the other hand, whether bot or human, were all positive.
“[T]hey generate almost no negative tweets,” Bessi and Ferrara write of the Trump bots. “They indeed produce the most positive set of tweets in the entire dataset — a very significant fraction of these non-negative bot-generated tweets (about 200,000 or nearly two-third of the total) are in support of Donald Trump.”
The two researchers also discovered that while bots were less effective at getting humans to engage with them than human Twitter users were, there wasn’t a significant difference when it came to generating retweets. They chalk this up to something intuitive: Bots may not be complex or sophisticated enough to get into conversations with, but people will retweet them if it helps to make a point.
There isn’t a clear solution to the constant presence of bots on Twitter, since the company has a lot of trouble culling the malicious, spammy bots, and is probably more focused on the fact that nobody wants to buy it right now. And while Bessi and Ferrara note that the study they conducted was relatively short, they’re concerned about the fact it’s hard to tell where the bots come from, who’s running them, and other information about them that’s pertinent to understanding how these bot armies are trying to influence political discussion.
They conclude that “future efforts, will be required by the machine learning research community to develop more sophisticated detection techniques capable of unmasking the puppet masters.”
In the meantime, hope you can figure out which Twitter accounts are puppets.