A third of Twitter users tweeting about Alberta politics might not even be real people.
That’s the findings of Mentionmapp Analytics — a Canada-based social media investigation and analysis company — in a report commissioned by the Alberta Federation of Labour.
The study looked at activity between Jan. 23 and Feb. 29, 2019, in which 2,995 Twitter profiles tweeted using the hashtags #abpoli and #ableg. Looking at accounts with suspicious activity — namely a higher than average number of tweets — Mentionapp found that nearly 29 per cent of those profiles were very likely to be bots.
A 2017 study from University of Southern California and Indiana University estimated 15 per cent of all Twitter accounts are bots.
The Mentionmapp study found that 361 of the profiles tweeted an average of 179 times per day, or one tweet every eight minutes.
A further 474 profiles tweeted an average of 74 times per day, or one tweet every 19 minutes and 27 seconds.
The report compared the Twitter output of ordinary people like AFL president Gil McGowan (17 times per day) and media outlets like Global Edmonton (45 times per day) over the same amount of time.
“I’m not sure what we see here is an anomaly,” Mentionmapp CEO John Gray told Global News.
“It doesn’t surprise me at all,” said Tom Keenan, University of Calgary professor and author of Technocreep. “Bots are really easy to do, basically free to create. So why wouldn’t somebody who wants to get their opinion out there use them? It’s almost a no-brainer that people will create bots.”
Twitter bots will insert themselves into conversation threads and strongly argue a political point, often from either the left or right of the political spectrum.
“Just today I read a paper on the neuroscience of marketing that talks about certain phrases that excite people — generally negative ones, ones that inspire fear because people are very motivated by fear,” Keenan told Global News.
“The Alberta election is important, so of course people are going to use every tool that they can. The problem is, unscrupulous people are going to have a lot of influence.”
Gray said the study can’t prove the impact of the activity or show who the actors are behind the Twitter bots.
“What we can conclude is that there’s questionable behaviour of ill intent to manipulate and influence.”
How to spot a bot
Keenan pointed to cybersecurity company Symantec’s guide on how to detect Twitter bots.
“There are bot creation engines, and that’s why sometimes you’ll see numbers as part of somebody’s Twitter handles,” Keenan told Global News.
“Another thing to look at is do they have a lot of followers or do they have a lot of activity.”
And while some bots are recently-created, not all bots were created in 2019, Keenan said.
“One of the most interesting things is there are actually companies that create Twitter accounts and age them like fine wine so, when they actually get used, people will look at and see the account has been around for three years. But that doesn’t mean they’re real.”
The effect of bots on Twitter
For Gray, who has studied the number of bots involved in Twitter discussions of B.C. and Ontario politics, the existence of Twitter bots isn’t the concern. It’s the narrative the bots are pushing that is of concern.
“What does it do, if you see a meme or narrative often enough, does that start to seem normal? If people pour enough of it into the stream, does the average person start to believe this is normal, this is true, this is reality? Does it drive people from wanting to participate in the conversation?”
“I think it should be cause for concern not just for folks in Alberta,” Gray said. “I have graver concerns that this is just a precursor for what’s coming down the pipe federally, what’s going to transpire in the fall.”
Symantec’s guide on how to spot a Twitter bot:
- Many Twitter bots have a relatively recent creation date.
- Many bot user names contain numbers, which can indicate automatic name generation.
- The account primarily retweets content, rather than tweeting original content.
- The account’s tweet frequency is higher than a human user could feasibly achieve.
- The account may have a high number of followers and also be following a lot of accounts; conversely, some bot accounts are identifiable because they send a lot of tweets but only have a few followers.
- Many bots tweet the same content as other users at roughly the same time.
- Short replies to other tweets can also indicate automated behavior.
- There is often no biography, or indeed a photo, associated with bot Twitter accounts.