Finns were reluctant to share election tweets of bots

According to the final report of ELEBOT project, the effect of Twitter bots was minimal during this year’s Finnish elections
Kuvituskuva, jonka pohjana käytetty visualisointia eduskuntavaaleihin liittyvistä aihetunnisteista

The number of Twitter bots and their influence on the Finnish elections this spring were minimal. Researchers make this conclusion in the final report of ELEBOT project. Despite the small number of bots, the results indicate that bots distribute unequally over the internet.

Bots produce either real or distorted information in an automatized manner. They are one form of information operations that aim to influence on, for example, election results. Aalto University researchers found information operations in content shared on Finnish Twitter before the Finnish parliamentary and European elections. However, bot tweets received only little attention from normal Twitter users who seemed reluctant to retweet content created by bots.

Before the parliamentary election, a bot wrote only half percent of all retweets by normal users and 0.7% of retweets were written by a so-called stopped account. Stopped accounts are user accounts that Twitter has closed or frozen due to spam or another security risk. Before the European elections, equivalent proportions stayed below half percentage.

The difference is large in comparison to, for example, the US mid elections held in November 2018. In the US elections, more than 35 percent of bots’ tweets were retweeted at least once. Researchers of ELEBOT project used same measures that have been used, for example, in the US elections.

The distribution of bots is unequal

According to the report results, bots tweeted mainly about the same topics as normal users. However, bots and normal users retweeted different types of tweets and interacted with different users. This suggests that bots’ operate unequally over the internet.

By using the most popular retweeted topics, the researchers were able to visualize communities whose members were more linked to their in-group members than users of their out-group.

The researchers recognized the three largest communities that were visualized in green, purple, or blue in the report. The smallest community was the blue one, formed by about seven percent of all users. In it, the relative share of certain hashtags – often relating to immigration or the right wing populist Finns Party – was larger than in the other communities. Nearly eight percent of bots and stopped accounts belonged to the blue community.

In the European elections, the distribution of bots and stopped accounts was similar to that in the parliamentary election. In the blue community, less than 2% users were bots, but they interacted with each other quite often; around 14% of all interactions between bots and stopped accounts happened in this community. Based on visual observations, such interactions were more frequent in the blue community than in other communities.

Suurimmat yhteisöt visualisoituna. Mukana myös niiden käytetyimmät aihetunnisteet.
The large picture visualizes the three largest communities by using green, purple, and blue colors. The most shared hashtags of the communities whose tweets were shared most often are included in the picture (in Finnish).

Twitter isn’t the only platform for information operations

The data gathered for the parliamentary election analyses included 188,668 unique users who published more than 1.3 million tweets in total. For the European elections analyses, the number of tweets in Finnish was 490,000, and 46,000 users tweeted about the elections. When more global topics were taken into account, the number of relevant tweets increased to nearly 660,000 and the number of Twitter users around 118,000.

About five and half percent of the Twitter users who tweeted about election themes before the parliamentary election were bots. They created 2.5% of all election-related tweets. During the EU elections, around 2.8% of all users tweeting about the elections were bots, producing 1.8% of all election-related tweets.

Even though the number of bots and their influence was minimal, researchers point out that the model used for bot-recognition may be unable to recognize all bots. Bots progress quickly and creating a better model calls for deeper development work. It is also important to remember that a normal user who isn’t classified as a bot can try to influence public discussions in a similar manner as a bot.

Information operations may have occurred also in other social media channels. In principle, Twitter is an open platform and the data collected by it is easy to access. The extent to which bots had effect on elections would be best discovered by taking into account also other platforms, such as Facebook. However, many services are more difficult to research as they tend to less open than Twitter.

Researchers gathered data by using the Twitter API from January 20 to April 15 for the parliamentary elections. They studied the tweets related to the EU elections from April 14 to May 31.

The report published by Aalto University has been written as part of the ELEBOT project that has been funded by the Finnish Ministry of Justice and Ministry of Education and Culture. In addition to the funding parties, the steering group of the project included representatives from the Finnish Security Committee and Prime Minister’s Office.

Link to the research report (in Finnish): https://aaltodoc.aalto.fi/handle/123456789/39200

 

Further information
Mikko Kivelä
Assistant Professor, Department of Computer Science
Phone 050 3824 555
[email protected]

Raimo Kantola
Professor, Department of Communications and Networking
Phone 040 7501 636
[email protected]

 

Further information about the report content and analytics
Ali Salloum
Research Assistant, Department of Computer Science
[email protected]

Tuomas Takko
Research Assistant, Department of Computer Science
[email protected]

  • Published:
  • Updated:
Share
URL copied!

Related news

Fresh ideas-drupal
Press releases, Studies Published:

Apply to Aalto Thesis project: Population Register Centre – Design System in Digital Public Sector

Aalto Thesis programme’s application period for master’s students is open. Read more about the Population Register Centre’s project and apply now!
An illustration of a ray of light causing the eye to find a route through a maze in the brain. Illustration by Safa Hovinen.
Press releases, Research & Art Published:

Studying vision in pitch-darkness shines light on how a mammal’s brain drives behaviour

Neuroscientists link mammalian behaviour to its underlying neural code at the unprecedented resolution of individual nerve impulses for the first time.
3 red letters of H and text "Finland".
Press releases Published:

Aalto Online Learning nominated for Hacking Higher Education finals!

Aalto Online Learning has been nominated to be one of the Hacking Higher Education Finland finalists! The finals are held in Dare to Learn learning festival at Korjaamo on Friday 20 September at 14.
YLE-yhteistyö
Press releases Published:

Yle aims to recruit top talent from the technology and creative sectors and will join forces with Aalto University to expand their cooperation

Yle and Aalto University have reached an agreement on communicating about jobs, traineeships and exciting career paths for the students of Aalto. The Employer Services Key Customer Package will be valid for one year as a first step.