Behind the bot: Misinformation online

x Sponsored content on Research Live and in Impact magazine is editorially independent.
Find out more about advertising and sponsorship.

Think back to the morning of 24 June 2016. Many people were in shock, having woken up to the news that the UK had voted to leave the European Union. For them, the result went against all expectations and, in the years afterwards, a reason for the outcome was sought. Some believed bots – automated accounts present on social media sites such as Twitter – were to blame for distorting the vote.

There is a lot of evidence that bots were present in the social media debate surrounding the EU referendum. In 2017, a University of London study by Dr Marco Bastos and Dr Dan Mercea found 13,500 Twitter accounts were tweeting extensively about the Brexit referendum, only to disappear soon after the vote. The bots had posted 65,000 messages over a four-week period, which the researchers said were slanted in favour of the leave campaign, albeit with many messages also pro-remain.

Bots have since been observed at other major political and social events, such as the 2016 US presidential election and the initial wave of Covid-19 in 2020. Bots have also been blamed for the flood of misinformation that swirls around many online debates and associated rises in conspiracy theories.

Do bots deserve their bad reputation? Can they actually turn election results or cause people to adopt extremist views and political stances? University of Edinburgh researchers Abeer Aldayel and Walid Magdy investigated the role bots play in the spread of misinformation in their study Characterising the role of bots in polarised stance on social media.

Their study analysed a dataset of more than 4,000 Twitter users who expressed a stance on seven topics: Hillary Clinton; climate change; feminism; abortion rights; atheism; Brexit; and immigration. Those users’ direct interactions and indirect exposures with more than 19 million accounts were then investigated, with bot accounts identified.

The supporting/opposing stances were noted and compared with other types of accounts, such as those of influential and famous users.
The researchers looked at two types of network: interaction and exposure. The interaction network is, broadly, the accounts with which a user retweets and interacts. The exposure network is the ones they follow and to which they are exposed. The results found that, generally, bots were far less effective at spreading misinformation than the accounts of influential users.

Bot interactions with users who had specific opinions or political stances were minimal compared with influential accounts with large numbers of followers. However, the presence of bots was still connected to users’ views, especially in an indirect manner – users were exposed to the content of the bots they follow, rather than by directly interacting with them by retweeting, mentioning them or replying to tweets.

“We found some bots were influential, but they were really minimal – we are talking about 5% or 6%,” says Magdy. “Yes, there are bots; they might have some correlation with our opinions, but we are not seeing them influencing it.

“A tweet from someone like Donald Trump or Barack Obama will be more than 10 times more influential on your opinion compared with bots. Yes, bots exist and have some influence, but it is really minimal.”

There is also a misunderstanding of what bots actually are and the role they can play online. Unlike many other social media platforms, Twitter allows accounts to post tweets automatically, which has led to many bots being created, a minority of which are used in harmful ways.

Most are used positively, however; for example, Wikipedia runs an @EarthquakeBot account that gives updates on earthquakes above five on the Richter Scale. Twitter also has rules in place to prevent automated accounts from spamming users or sending unsolicited messages.

Magdy says there is a fear of bots that is generally not supported by their actual impact on public discourse. “That fear might be overrated. Focus on the influential users,” he explains. “It is like shouting a lot in a closed room – nobody hears you much and it does not have much effect.”
Some social media platforms, such as Facebook, do not allow bots (although Facebook does have an issue with fake accounts).

“Each [network] has its own nature,” Magdy says, adding that Twitter is one of the most popular platforms for news dissemination, making the spread of misinformation on the site more worrying. “When we talk about misinformation and stances about any topic in general,” he says, “we noticed bots are there but don’t have much influence.”

The problem with tackling bots is determining which are malign actors and which remain useful. There is also a big spectrum of what constitutes misinformation. There are various shades of grey between the outright falsehoods and the complete truth – for example, using cherry-picked statistics or false equivalences. Magdy is working on a project to understand why some people do not check information they see online and the common characteristics between them: “Disrupting the sources [of misinformation] would be great, but can we find a way to stop the consumption as well?”

Even in the context of Brexit, however, Magdy questions how much of an impact bots had on the result. “Even if it is 30% [of accounts on a social media network that are bots], they still don’t have much impact,” he maintains.

“Yeah, there are bots everywhere. The media amplified it and made it a big problem, and said that it changed votes – but that’s too big a conclusion.”

This article is from the July 2022 edition of Impact magazine.