Political bots spread misinformation during U.S. campaign

As usual, many Americans (and Canadians) have turned to social media for analysis and commentary during this year's U.S. election.

But much of the online conversation hasn't come from humans at all. Instead, a significant number of social media posts are generated by robots.

CBC Radio technology columnist Dan Misener looks at the rise of "political bots."

What is a political bot?

A political bot is a program — a piece of software — that usually operates on a social media website like Twitter or Facebook. Normally, they automatically generate social media posts, which may look like they've come from a human being.

The goal of a political bot in particular is to promote a specific ideology or public policy idea. For instance, there are armies of pro-Trump bots, and pro-Clinton bots, and they've been very active throughout the 2016 presidential election campaign.

• Partisan Twitter bots distort candidates' popularity
• Day 6 | Can Twitter taunt bot defeat the trolls?

Phil Howard, who studies bots — and particularly political bots — with the Oxford Internet Institute estimates that a third of the millions of tweets in favour of Donald Trump, and a quarter of pro-Clinton tweets, are generated by automated accounts.

Some of them may seem like they came from legitimate accounts. But there are some ways to spot a bot, Howard says.

"A bot is usually easy to identify because it's got a crazy number of followers, hundreds of thousands of followers, or it follows hundreds of thousands of other people, and you. Sometimes a bot has sent one message in the last year," he told CBC's Spark. 

"Usually a bot also has also no photos — has the little egg [profile picture], say, for Twitter. It's usually easy to identify them, because they don't seem to have a real social network behind them."

• VIDEO: Spot the bot

But these political bots often don't identify themselves as bots. They pretend to be human users. And they're often used for negative campaigning, or to spread misinformation. They will target specific users and harass them, intimidate them, or try to choke off a conversation.

Of course, there are also "good bots" that sometimes perform a public service. But generally speaking, we're talking about large armies of automated social media accounts used to manipulate public opinion and spread misinformation. That's why some researchers call them "computational propaganda."

What kind of misinformation do they spread?

Howard has watched how political bots have been used throughout the current American election campaign and highlighted some of the misinformation he's seen.

"The most recent one we're tracking now is a set of tweets about how Americans shouldn't vote — that voting day has been postponed until Thursday, and that they can vote for Hillary [Clinton] by SMS — by texting a certain phone number," he said.

• Bot creates 'remarkably Trump-like' tweets

"Now, most voters wouldn't take that stuff seriously. But if they can get a few percentage points of the electorate to think that voting has been postponed until Thursday, that'll have consequences."

The bots don't actually create these lies. But they spread them, repeat them and perpetuate them.

How much bot activity have we seen?

These bots have been busier and busier as the campaign has progressed.

Howard and his colleagues looked at bot activity during each of the three presidential debates. And they found bot activity blossomed after every debate. At its peak, about one in three tweets about the debate was generated by a bot.

Also, the ratio of pro-Trump bots to pro-Clinton bots changed over time. For the first two debates, Clinton bots were outnumbered with four pro-Trump bots for every pro-Clinton bot.

By the final debate, that gap had widened — there were seven pro-Trump bots for every pro-Clinton bot. 

The timing of the bot activity also changed over the campaign. In the first two debates, bot activity roughly mirrored human activity. Bots and humans were tweeting at the same time — during and after the debate.

The innovations that happen, that evolve out of the U.S. election, carry over into other democracies.- Phil Howard,  Oxford Internet Institute

But by the third debate, the bots got a head start and started ramping up their automated messages in the hours before the broadcast started. That suggests there is human strategy behind these large armies of bots.

Why is it important to understand political bots?

Howard says the techniques we see developed during a big-money U.S. election eventually get exported to other countries.

"The political consultants who work [on U.S. campaigns] go off to London and Ottawa and Canberra, and they ply their trade there," he said.

"The innovations that happen, that evolve out of the U.S. election, carry over into other democracies. Very soon we'll see the same kind of techniques in the next Canadian election. We've already seen them in other democracies."

How effective are political bots?

Right now, we can't know for certain how influential political bots are. It's very difficult to connect online bot activity to changes in public opinion or voter turnout.

Of course, researchers like Howard — and politicians themselves — are very interested in figuring out the return on investment of bots. If a campaign decides to spend money on an army of political bots, is it worth it?

Howard says he expects we'll see more political bots in the future, and they'll be increasingly sophisticated.

And that may be the most worrying part, because it's not always clear who's a bot. One that's well-programmed can do a reasonably good job of passing itself off as human.

When it comes to the U.S. election, the impact of bots will be difficult to measure. But if there's any indication that they could be an effective campaign tool, expect to see even more of them.