Devoted users: EU elections and gamification on Twitter
By Francesco Grisolia and Antonio Martella
Launch of “Vinci Salvini!”, May 5th 2019 (screenshot of the official video). The related post (right) briefly describes the rules of the contest.
Our study, whose preliminary results we recently presented at the 2019 SISP (Italian Political Science Association) Conference, examines the visibility of the tweets posted by Italian political leaders during the last EU election campaign. We show how crowd-sourced and spontaneous political action, triggered by a social media game, can take an almost social bot-like nature and significantly boost the visibility of tweets by political leaders during a major political event.
In particular, our analyses on the Italian political Twittersphere show the clear predominance of Matteo Salvini, former Italian Interior Minister and leader of the Lega (“League”) party. Salvini’s intense activity is unparalleled by any other leading Italian figure; for this reason, we decided to focus on his most retweeted tweets in the month before the elections.
The figure below shows the daily amount of retweeted tweets of the five main Italian political leaders: Silvio Berlusconi, Luigi Di Maio, Giorgia Meloni, Matteo Salvini, and Nicola Zingaretti. Matteo Salvini is represented by the green line, which is clearly above the others.
Daily amount of the main Italian political leaders’ retweeted tweets (April 26th - June 1st 2019).
While political messages travel across a platform, they can be echoed by different types of users – human, automated (bot) or in between (cyborg) – with varying degrees of legitimate or malicious behavior. In our study we decided to apply the Digital DNA toolbox because its approach has been proven particularly effective, performing better than other state-of-the-art algorithms.
Gamification is widely conceived as the use of video game elements in non-gaming domains, such as work or healthcare. In recent years, such an approach has been applied to the realm of politics through social media campaigns, with the intended goal of increasing participation and engagement.
Read more about the Digital DNA Toolbox
An interview with the Digital DNA Toolbox team
How to map our online behavior
The social media contest that we examine in our study, “Vinci Salvini!”, made its first appearance during the 2018 Italian general election campaign, and whose second edition was launched three weeks before the 2019 EU Elections. In order to enter the game, any contender had to have a Facebook account, through which s/he had to register to the contest website. The contenders were asked to share personal data, email address and mobile number. They could also consider adding their Instagram and Twitter accounts to increase their chances of victory. The competition was quite simple: they had to like posts on Facebook and Instagram, and to like and retweet tweets posted by Matteo Salvini’s official accounts as much as they could, and as quickly as possible. It is worth noticing that the speed of their retweeting activity of any content could increase its chance to become a trending topic on Twitter. Daily winners were to be awarded a call by Salvini himself and a post with their pictures, to be shared through Salvini’s official social media accounts. On the other hand, weekly winners were to be awarded a private meeting with Salvini and, had they wished, a video to be posted on Facebook, Instagram, Twitter, and Youtube. Additionally, all the winners were to receive a pass for Salvini’s closing rally in Milan on May 18th.
Launch of “Vinci Salvini!”, May 5th 2019 (screenshot of the official video). The related post (right) briefly describes the rules of the contest.
We collected our data using the standard Streaming and Rest API. In total we gathered 463,886 retweeted tweets posted by Matteo Salvini’s official account (@matteosalvinimi), relative to 30,853 unique users.
We focused on his five most retweeted tweets and we gathered the timelines of the users who retweeted them. Then we coded the timelines according to the actions (tweet, retweet and reply) performed, and compared them using the Digital DNA technique.
This stage of our research showed the presence of many users sharing long LCSs (Longest Common Substrings), i.e. sequences of identical actions (tweets, retweets or replies), but it did not unequivocally show the operation of social bots or botnets.
As a following step, and in order to clarify if the gamification strategy promoted by the leader of Lega affected users’ behaviors, we analyzed the temporal patterns of their retweeting activity. On a manual inspection of these profiles, we noticed that most of them only retweeted the tweets posted by Salvini during the last month of the election campaign, irrespective of the age of their accounts. We decided to isolate those users whose timelines were mostly composed of Salvini’s retweeted tweets, according to the formula:
Amount of Salvini’s RTs during the monitoring period >= total statuses / 2
Quite surprisingly, just the 2% of users were responsible for more than the 26% of the total amount of Salvini’s retweeted tweets during the last month of the election campaign.
Plotting the activity of all Matteo Salvini’s retweeters, we found that selected accounts (figure 3, area in red) started their activity just after the launch of “VinciSalvini!” (May 5th) and strongly reduced it just the day before the elections (May 26th). Such temporal pattern supports the hypothesis that Salvini’s gamification strategy has been particularly effective, increasing the contenders’ engagement and in turn the visibility of his tweets.
Daily amount of retweeted Matteo Salvini’s tweets
We then examined the characteristics of the users whose activity on Twitter seem to have been triggered by the social media contest. For their intense retweeting activity we decided to label such accounts as devotees, suspicious accounts simultaneously showing a bot-like behavior and distinctive human features.
According to the Oxford English Dictionary, devotion is “love, loyalty, or enthusiasm for a person or activity”. The Collins Dictionary adds “commitment” to this definition. Enthusiasm and commitment to a cause emerged as the characteristics of these accounts. When such profiles emerged from our analyses we first reacted with puzzlement: they seem almost indistinguishable from social bots in terms of online behavior, but they also appear to be controlled by humans.
The devotees tend to share four features:
Account age
Temporal dynamics of online behavior
Type of activities
Amount and type of friends
First, they generally are recent accounts; many have been created in the run-up to the elections, specifically a few days before or after the start of the “Vinci Salvini!” game. Analogously, the accounts created before May 2019 significantly increased their activities just after the launch of the “Vinci Salvini!” contest. Third, their activity on Twitter is nearly limited to retweeting and liking Salvini’s tweets. Finally, most of them have a limited number of followers and an even smaller amount of friends: many exclusively follow Salvini’s official account, while a larger number of devotees also follow the Lega party and Giorgia Meloni’s official accounts.
They look like accounts exclusively created for one purpose: to follow, like and retweet Salvini’s and Lega’s content. In almost any aspect of their online behavior they suspiciously look like automated accounts, yet on a manual inspection they reveal a distinctive human aspect: the type of replies that they post occasionally are too long and articulated to be algorithmically generated. Differently from cyber troops or trolls, there are no elements to judge them as state-sponsored agents, engaged in coordinated efforts of political manipulation. They seem to spontaneously act in support of a political leader, not driven by monetary incentives (as click workers), but by political affinity. Finally, they seem to keep complete control of their own Twitter accounts, so they cannot be categorized as cyborgs. In light of all these peculiarities, we see the devotees as a new form of crowdsourced political agents, triggered by gamification processes.
About
Francesco Grisolia received a PhD in Social Anthropology from the University of Siena (Italy) and is currently a postdoc fellow at the University of Pisa. He does research on online disinformation, with a focus on social media, political parties and leaders. He is a member of the KDD Lab (ISTI-CNR, Italy) and MediaLab, a joint research group (University of Pisa - CNR) investigating big data in social and political research. Email: francesco.grisolia[at]di[dot]unipi[dot]it
Antonio Martella, PhD, is a post-doc researcher at the Institute of Informatics and Telematics of the National Research Council of Pisa. He is a member of the MediaLaB, Big Data in Social and Political Research Laboratory of the University of Pisa. He has earned a postgraduate master's in Big Data & Social Mining at the Computer Science Department of the University of Pisa. His research interests focus on populism, leaders and political communication on social media, analyzed through big data and statistical methodologies. Email: antonio[dot']martella[at]iit[dot]cnr[dot]it. Twitter: @vot4ntonio ORCID: 0000-0003-3378-1782.