The number of Twitter bots supporting the anti-immigration right-wing Sweden Democrats has doubled in the build up to the Swedish election on 9 September.
The report, which was commissioned by the Swedish Defence Research Agency, looked at nearly 600,000 tweets which featured hashtags associated with the election and found up to 17 per cent of them came from automated accounts.
It found that these Twitter bots were 40 per cent more likely to show support for the Sweden Democrats compared to genuine accounts.
Yougov polls suggest that the right-wing party, led by Jimmie Akesson, will receive the highest percentage of the vote with the current poll placing them on 24.8 per cent – one percentage point higher than the Social Democrats which currently lead government.
Ralph Schroeder, who worked on the report, told NS Tech: “We used machine learning to identify any activity which, according to the model, can be identified as automated behaviour – so bots and humans who are acting as if they are bots in order to spread disinformation.”
They found that 47 per cent of the Twitter bots supported the Sweden Democrats whereas among genuine accounts they were only supported 28 per cent of the time.
The report also found Alternative for Sweden, which hold views even further to the right of the Sweden Democrats, were discussed by 29 per cent of the bots. Schroeder described these results as “very disproportionate”.
The most common topics discussed by these automated accounts were criticism of elites, media criticism and criticism of immigration policy and Schroeder warned that spreading disinformation had become “a factor to be reckoned with” and “should be a concern for democratic discourse around an election”.
Speaking on the reason they chose to focus the study on Twitter he said: “It is public, and it’s been shown that journalists and people with an interest in politics are quite active on Twitter, so Twitter is a sphere of public discourse which is rather important and gives you a different picture to traditional media.”
Schroeder added: “The motivation is clear, it is trying to skew or support forces by means that are not genuine. Whether it will affect the upcoming election is hard to say.
“To put a positive spin on it, I think people will be watching like hawks what happens at the upcoming Swedish elections and American midterms. Teams will be looking at Russian, foreign and bot activity in the US and it will put the issue under the microscope.”
Twitter responded claiming that the research was working off their API which “does not take into consideration the host of pre-emptive strikes we take to ensure that automated content is not visible in search, in trends, and in all discoverable areas of the service”.
They added that in May their systems identified and challenged more than 9.9 million potentially automated accounts per week.