Are bots a danger for political election campaigns?

FAU researchers have been investigating how computer programmes can be used for propaganda purposes

Normally, autonomous computer programmes known as bots are used to trawl the Internet, for example to help search engines. However, there are also programmes known as social bots which interfere in social media, automatically generating replies or sharing content. They are currently suspected of being used to spread political propaganda. Scientists at FAU have investigated the extent to which autonomous programmes such as these were used on the platform Twitter during the general elections in Japan in 2014. By using methods taken from corpus linguistics, they were able to draw up a case study on the various activity patterns of social bots. At the same time, the FAU researchers gained an insight into how computer programmes like these were used and recognised that nationalistic tendencies had an important role to play in the election, especially in social media. The results of the investigation have been published in the journal Big Data (DOI: 10.1089/big.2017.0049).

Prof. Dr. Fabian Schäfer, Chair of Japanese Studies at FAU, was motivated to study the use of social bots after the general election in Japan in 2014. The election was won by the conservative Liberal Democratic Party of Japan, led by Shinzō Abe. Publicly and in the mass media, his election campaign focused predominantly on economic issues. It was a different story in social media. ‘Our analysis showed that Abe’s hidden nationalistic agenda had a very important role to play in these channels,’ Schäfer explains. ‘The importance of the hidden agenda in social media is not, however, down to either the prime minister or the LDP itself.’ Rather, it appears as if social bots were widely used by right-wing Internet users, ranging from far-right to more conservative right-wing circles. Prof. Schäfer’s initial hypothesis was that the right-wingers used social bots to give indirect online support to Abe’s nationalistic agenda, which had slipped into the background during the political campaign.

Computerised propaganda

Together with Prof. Dr. Stefan Evert, professor for Corpus Linguistics at FAU, Schäfer analysed over 540,000 Tweets posted shortly before and just after the elections in mid December. After noticing a high frequency of Tweets which were the same or very similar, they investigated whether these originated from bots or even botnets. Unlike previous studies, the FAU researchers did not focus on identifying bots based on their typical activity pattern, for example how often they send Tweets. Instead, they took a corpus linguistics approach, which allowed large volumes of text to be analysed. It quickly became clear that nearly 80 percent of the investigated Tweets were duplicates, including re-tweets, or close duplicates, which were able to be traced back to a total of 3722 original Tweets.

The scientists at FAU recognised five patterns behind the spread of the Tweets. They attributed three of the patterns to a pro-LDP campaign, and one to another group of right-wing Internet activists. The fifth pattern attributed to users who acted similarly to bots. ‘Tweets in the first and second groups use similar language, reflecting the jargon often used in right-wing Internet circles, and tend to include racist or hostile remarks,’ Schäfer explains. This and the names used for the numerous fake accounts led Evert and Schäfer to the conclusion that the Tweets originated from two groups of right-wing Internet activists, the netto uyo, who were making massive use of these automated programmes to cover other hashtags. Bots can also spring onto popular hashtags and Tweets to instrumentalise them for the same purpose, without the Tweets sharing the same content.

This was exploited by right-wing Internet activists in the election in 2014, in particular to boost the spread of extremely nationalistic content. The accusation of being ‘anti-Japanese’ – a term which is used both by netto uyo and in a slightly less extreme form by Abe – acted as a kind of language bridge between the extreme and the more moderate groups. According to Schäfer, ‘This bridge connected the nationalistic discourse of the right-wing Internet activists with Abe’s right-wing conservative agenda. As a result, Abe’s position was not only supported by the conservative organisations of a group of users with close links to the LDP but also by the large, although not well organised, group of right-wing Internet activists.’ As the Japanese expert Schäfter explains, even though this group often took an anti-Abe position, they were spreading a very similar nationalistic agenda online.

Original publication: Japan’s 2014 General Election: Political Bots, Right-wing Internet Activism, and Prime Minister Shinzō Abe’s Hidden Nationalist Agenda. Big Data. DOI: 10.1089/big.2017.0049

Further information:

Prof. Dr. Fabian Schäfer
Phone: +49 9131 8529145
fabian.schaefer@fau.de