Activists, Not Bots: An Analysis of #VoteLabour Tweeters

The surprise election result has produced a lot of early responses, which offer avenues for more detailed research.  Amongst these have been suggestions that Labour activist groups made effective use of social media campaigning.  Research conducted into the party’s support on social media, including some carried out by Demos, has suggested that pro-Labour Tweets generally outnumbered all other party-political Tweeting.

There have been claims that this high volume of pro-Labour Tweeting may have been boosted by automatic ‘bot’ accounts.  A report by the Political Bots project – a frequently cited source in reportage on social media usage during elections – suggested that pro-Labour Tweets included more automated traffic than Tweets supporting other parties, a finding which was reported on the front page of The Telegraph newspaper.

The possibility that automated accounts have been involved in generating large amounts of political content online is one we should take seriously. Since bots are able to post indefatigably and at high volumes, the presumed danger is that these tireless accounts could be employed to artificially inflate the support a party or candidate seems to have, or sling mountains of mud at opponents.

It is important, however, that we don’t write accounts off as being automated too lightly. To mislabel human Tweeters engaging in political discussion is to dismiss genuine democratic participation: there is no more absolute way of denying that a person’s views are legitimate than by denying that they are in fact a person. Due to the indications that social media played an important role in this election, we believe the appearance of high-frequency Tweeting from pro-Labour accounts, and their status as humans, merits further investigation.

Analysis of the #VoteLabour Tweeters

For this analysis we used Twitter’s public API to collect all Tweets sent worldwide between 2^nd and 8^th of June which contained a generic election hashtag (listed in full below.) The following analysis focuses on a single hashtag: #VoteLabour. This is the most commonly used partisan hashtag during the election and the second most common hashtag in our dataset, after #GE2017.

We looked at Tweets sent in the week leading up to and including the day of the election, a period which saw an intense increase in General Election related Tweeting.

The measure used by the Political Bots project across multiple elections is that any account which tweets over 50 times a day on a single hashtag is a bot.  Looking at the data on a day-by-day basis, we located 22 unique accounts who fitted this requirement.  Of these:

• 5 were official or unofficial Labour accounts, such as @local_labour or @manc_labour. The high rate of Tweeting of these accounts could be explained by a group of people using the same account.
• 11 were highly unlikely to be bots, as they sent Tweets which showed intelligent response to a specific context.
• 6 could have been bots, in that all of their Tweets could have theoretically been sent without any human intelligence. However, even these were occasionally doubtful.  Many sent Tweets on a variety of topics, rather than simply copy-pasting or retweeting material related to the Labour party.^[1]  Others had been created in 2012 or earlier,^[2] or provided easy-to-check personal contact details in their biography.  Overall, there was only one account that could be confidently labelled as a bot.

These results suggest we should be cautious of seeing the high volume of pro-Labour activity online as indicative of bots.  Instead, these high-frequency accounts suggest alternative explanations.

Of the 11 accounts which were run by real people, many were clearly activists for social justice more broadly; their Tweets supporting Labour appeared alongside other concerns such as protesting cuts to the NHS or supporting community responses to the Grenfell Tower fire.  Biographical details suggested other factors which could account for their high activity, such as being retired (allowing more time to Tweet) or having been a victim of disability cuts.

Various analyses of digital politics have shown a strong uptake and use of social media by activists for various causes.  This suggests a tentative early hypothesis, which merits further investigation: that the volume of pro-Labour digital campaigning, and any success that may have brought the party, could be due to the social justice message of Corbyn’s campaign appealing particularly to people who were already extremely active online campaigners.

Ongoing Work:

There are limitations to the approach we have taken.  Searching on a single hashtag means we may only capture a small proportion of an account’s Tweets – it would not distinguish between someone who Tweets at a perfectly human rate 50 times a day on that hashtag, and someone who Tweets at the inhuman rate of 5000 times a day but only uses the hashtag for 50 of them.

However, from an early inspection of a much larger database of General Election related Tweets we have seen very little evidence of inhuman levels of activity.  Looking at the Tweets on a day-by-day basis, we see very few accounts – usually between two and five, depending on the day – who Tweet more than once a minute, and those that do tend to explicitly label themselves as bots or collaborative accounts (for example @liveeudebate or @isthisab0t).  There are multiple accounts who Tweet extremely quickly, but this is rarely sustained over entire days.  Further inspection on an hour-by-hour basis may reveal factors underlying this behaviour; for example live-Tweeting of speeches, as other Demos analyses have suggested, such as prominent speeches by Jeremy Corbyn, correlate with spikes in Twitter activity.

The issue of finding political bots is still a live one, and in collaboration with the University of Sussex we are using the General Election database to develop new methods for locating ‘bot-nets’ (accounts which have a human level of activity, but are actually part of a group programmed to Tweet identical messages at similar times).  However, the question remains as to whether the high volume of pro-Labour Tweeting suggests new avenues of inquiry which move away from the question of bots:

– To what extent did the social justice agenda of the Labour campaign energise a new generation of digital activists?

• Were many of the most active pro-Labour accounts highly engaged in digital politics prior to this election, and simply incorporated the #VoteLabour message into their existing activism?
• Does the digital engagement of Labour supporters go beyond voluminous Tweeting, and into implications for future campaigns – has this election perhaps formed new online networks of expertise, linking Labour ground campaigners with experienced digital activists?

Investigating these questions could provide useful material for the complicated but vital project of understanding the relationships between offline and online political engagement.

Search terms used to collect Tweets

#ge2017

#generalelection

#ge17

general election

snap election

june 8th

8th june

early election

#pmqs

#pmq

#VoteLabour

#VoteConservative

Corbyn

Theresa May

Teresa May

Farron

Sturgeon

Nutall

Caroline Lucas

Jonathan Bartley

Labour Party

Conservative Party

Tory Party

Tories

Libdem

Liberal Democrats

SNP

UKIP

Green Party

Theresa_May

CarolineLucas

Jon_Bartley

^[1] This is a common use of bots, which is extremely easy to set up, is to copy or retweet Tweets which contain a certain word.

^[2] Another tell-tale sign of political bots is that they have been created in readiness for a particular political event.