Both online campaigns looked choreographed to perform a specific function: to produce the appearance of popularity. Automation enables campaigns to game metrics, an urgent task in the poll-obsessed world of contemporary politics. The media increasingly seek to integrate statistics into their analyses to lend credence to their predictions, with quantitative data fetishized more than ever in the 2015 electoral campaign coverage. One of the unlikely celebrities of the elections, as The New York Times reported, was a hobbyist statistician who produced data visualizations to analyse the elections as they unfolded, which he posted on his blog (Gilbert 2015). La nación, internationally renowned for its creative integration of data and data visualizations into its reportage, repeatedly used the social media analysis company Flowics to monitor campaign action and support on Twitter and Facebook.
The move to new hashtags by both campaign teams, with followers provided with a timeframe for activity on them, suggests some minimal awareness of how to game the twitter trending topics algorithm. The algorithm captures change in intensity rather than high volume. Soon after the campaign season officially closed, La nación announced in an online article that Macri had ‘won’ on Twitter; #YoCambio had emerged as the primary trending hashtag, with #MejorScioli apparently trailing in third place. The statement echoed the result published across media after the second presidential debate (see, for example, Las encuestas de los medios 2015 and Macri arrasó en todas las encuestas de Twitter sobre el #ArgentinaDebate 2015). This kind of traction seems to have been the ultimate ambition of both campaigns. It suggests the extent to which they both imagined voters to conform to a herd mentality, by which logic to become popular the candidates needed first to appear popular, with the statistics to confirm that image.
If this particular ambition unified the two campaigns, differences also abounded in their approaches and the online images that they ultimately projected. One distinction lies in the level of centralized coordination and professionalization that appear to have shaped the two campaign strategies. In his work on the organizational structures behind the 2012 Mitt Romney and Barack Obama presidential campaigns, Daniel Kreiss (2014, pp. 3–11) outlines their respective approaches to social media, particularly Twitter. Although the structures and functionalities of the social media teams in the Republican and Democratic campaigns differed from one another, Kreiss describes both as highly professionalized and integrated into the overall campaigns. The social media effort around the presidential debates in the Romney campaign was pre-planned and meticulously choreographed. The Obama social media campaign, contrastingly, was more flexible. Social media staff had a clear collective vision that enabled individuals to work autonomously and spontaneously. We do not have access to the internal workings of either party’s social media strategy in Argentina, but the image that our data concerning automation provides suggests differences in the Argentine parties from both of the US campaigns.
Particularly in the Scioli Twitter campaign, professionalization appears lacking. The campaign began slowly. Then, it careened towards bulk account creation. This approach can provide no artifice of organicity; automation or at least coordinated account creation is a statistical certainty. Most bots were deployed almost immediately after their creation, with little effort going into populating them with followers, a backstory, or a credible-looking timeline. Blending in with real accounts was not always a priority. Deployed in bursts, this style of automation manifests little attempt to control or influence the online space beyond certain critical junctures. To be sure, as both US campaign teams were aware, timing can be key (Jungherr, 2014). Critical moments, such as a debate, can be planned for: the appearance of instantaneity or spontaneity requires neither instantaneous production nor spontaneous performance. Yet, the pro-Scioli material feels improvised. Occasional spikes in output were produced by individual accounts, likely operated through homemade scripts that tweet in large quantities. The sudden and sizeable ramp-up that characterized the second burst of automation suggests campaign staff learned on the job and perhaps indicates confusion surrounding best strategy. It is at least ironic that Carlos Gianella, technical director to Scioli, described the fake Twitters accounts that he attributed to the Cambiemos campaign as a sign of Macri’s ‘desperation’ (El Sciolista que denuncia 2015).
The activity surrounding the pro-Macri hashtags hints that more experienced hands were at work. The campaign was subtler and seemingly more professionalized. Both teams used automation to promulgate their message as widely as possible, but in Scioli’s case, verbatim copies of messages were pushed directly by supportive accounts. Both campaigns employed retweet bots, an approach that uses third-party bots for hire to increase the reach and visibility of material with relative subtlety. Like Scioli and his Ola Naranja, Cambiemos also apparently sought to blur the line between automation and activist. They equipped supporters with the tools for the decentralized implementation of automation, for instance by operating multiple accounts and the provision of downloadable images. As a result, the Macri campaign blended automation—centralized or produced by independent activists—with genuine grassroots activity and transparent official campaigning. If inconspicuous integration was the ambition, it was successfully met.
The question of expertise extends beyond campaign teams and their advisors alone. By the end of the runoff, Jorge Di Lello, the electoral prosecutor, had received 63 complaints, of which 52 lambasted President Fernández, for apparent continuities in electioneering after the embargo on campaigning began on 20 November (El fiscal recibió 53 denuncias contra Cristina por violar la veda electoral 2015). Di Lello announced that, among his investigations, he would analyze the President’s tweets to check if they violated the prohibition on campaigning. Such an approach appears superficial. It risks missing the main, and subtler, event—the use of automation after close of play. It is not only campaigning practitioners who must learn the art of online persuasion and manipulation, then, but the watchdogs too. Since the 1990s, the field of transparency and anti-corruption has consolidated and professionalized in Argentina, gaining public prominence and respect. As social media enables new forms of subversion, that field will need to adapt accordingly. Begun as a project by leading Argentine legal minds, our data suggests that it will need to acquire a new technical dimension: the watchdog can no longer only be a constitutional expert but must also be a network specialist, equipped to reveal covert structural manipulations online. The public attention that cyber activism has generated in Argentina hints at a recognition that Twitter is unregulated space and the mechanisms and knowledge to control it not yet fully established.
Our findings do not, of course, tell the full story about Argentine electoral campaigning on Twitter. For a start, though we find little evidence of negative campaigning via automation, it may have occurred through other, subtler, means. Smearing does not require automation: people, not just bots, love sensational news. What is clear, however, is that little Russian-style hashtag-capture took place. There has been no attempt to shut down spaces of information and opinion through streams of spam or irrelevant content. By and large, ‘spam-like’ material has been directed at candidates’ own information space. It is important to emphasize that our data form only a small portion of the Twitter commentary on the elections. Because the lens through which we approach the question of automation centres on hashtags, our focus has been on the areas to which the campaigns dedicated most attention. Popular hashtags will have acted as magnets for automated activity, certainly from spam marketers, but also likely from other bots that ‘piggy-back’ on trending subjects. The same levels of automation are unlikely to be observed in the ‘conversation’ at large. Limitations in our dataset mean we are unable to measure quantitatively the degree to which automated content in political campaigns is restricted to hashtags. More work in this area is need.
The online behaviours that we find here suggest that the elections have not triggered political communications strategies on Twitter that veer significantly from those surrounding other Argentine political events. In 2012, Kreiss identified how little we know about the ways in which campaign strategists and political elites ‘seek to channel, steer, influence, respond to, or otherwise manage networked political communication on social media platforms’ (Kreiss, p. 2). Our research suggests that this may differ at least as much from one political context to another as from one type of political event to another. During the 2011 Russian presidential elections, extensive hashtag flooding drowned out opposition voices (Elder, 2011). In a comparative analysis of political discourse on Twitter following the death of Alberto Nisman in Argentina and the murder of Boris Nemtsov in Russia, we detected similar behaviour on the Russian Twittersphere in the aftermath of Nemtsov’s death (Filer and Fredheim, 2016). In Russia, automation was used to hijack hashtags associated with the opposition, flooding them with spam to such an extent that it became almost impossible to stumble across real oppositional content. Bot clusters also tweeted at other bots, tweeted en masse ‘at’ individuals who wrote about #Nemtsov and specialized in retweeting high-profile bloggers. Automation undermined the potential of hashtags to expose individuals to oppositional material.
Following the death of Nisman, the emphasis in the Argentine Twittersphere was on amplification rather than drowning out opposing voices. Little use of automation was detected (Filer and Fredheim 2016). Twitter activity during the elections looked consistent among both camps with this ambition to promote messages and perhaps mobilize offline support, rather than to shut down alternative opinions. The principal novelty during the elections was methodological: the use of automation for amplification. The two teams seemed to perceive amplification on Twitter as a worthwhile investment, albeit only in bursts around key events rather than via the protracted automation seen in Russia. This turn to automation both changed the scale of political discourse on Twitter and created less of a genuine conversation. There were fewer uniquely worded messages, with retweets and duplicated messages—the result of automation or manual copy and pasting—dominating the space. Endless replication during the elections replaced the more authentic feel of the Twitter discourse around #Nisman. Taking two crude indicators of automation, proportion of duplicated content and proportion of accounts that tweeted about the hashtags within 10 days of account creation, the leap in levels of automation is clear. Automation on #Cambiemos was at least twice that on #Nisman, while for #ScioliPresidente and #MejorScioli we found between 8 and 10 times as much automation. The increase in scale and the reduction in polyvocal activity on Twitter do not alter the status of amplification as the end goal of political commentary on the Argentine Twittersphere. Read together with analysis of the Twittersphere following the Nisman affair, our findings suggest that political events beget social media responses that reflect the political environment in which they occur more than the type of event itself. Further research would do well to foreground the role of the domestic political ecosystem in determining the types of national discourses and strategies played out in the ‘global’ Twittersphere.
Attentiveness to national politics may also help to understand the role that accusations of automation played in these elections. If allegations of dirty campaigning have long abounded in Argentine electoral campaigning, why did the FpV focus now on automation? To have traction, the accusation required introducing the public to a new vocabulary—‘bots’, ‘automated content’, ‘botnets’—that required explanation. This complexity risked distracting from the straightforward and well-rehearsed task of painting the opposition as untrustworthy. One explanation is that the FpV’s claim was true, in outline if not in detail. The focus perhaps also diverted attention from Macri’s admonishment of negative campaigning by the Scioli camp, which included a televised advertisement comparing PRO economic policy to that of the military authoritarians of the 1970s and 1980s and another asking the viewer to ‘imagine the hunger’ that Macri’s presidency would beget (Argentina: Macri acusa 2015).
Twitter’s global reach and popularity may offer a further explanation of why the FpV highlighted how the medium could be used to deceive. Peronists have long observed disparagingly that opposing candidates hire and learn from foreign campaign strategists. The accusation serves as shorthand for selling the nation out, an omen of what the national future would hold should their opposition win (e.g. see Menem 1989, p.35). In the 2015 campaign, FpV supporters noted how Macri’s campaign team embraced strategies previously used abroad (Girondo 2015). By focusing on automation, the FpV did not so much depart from Peronist tradition as shape its latest iteration.
In the same way as accusations of unfair play pre-date the social media age, the practice of producing inflated images of popular support has long pervaded Argentine political life. Offline, examples and accusations of Argentine politicians and parties magnifying the appearance of their popularity are commonplace. These tendencies should perhaps be expected in a country that fuses electoral democracy and populist leadership. Populism frames the leader as the ultimate expression of the will of the people; the image of mass support is therefore intrinsic to the appearance of legitimate rule or serious opposition. Denunciations claiming crowds are ‘bussed in’ to inflate the image of popular support regularly accompany protests and rallies.Footnote 7 The limited use of automation surrounding the Nisman case is perhaps the bigger surprise than its presence within the electoral context. In a political culture fixated on the appearance of popularity, the use of automation to simulate mass support appears an organic development. So too does the attempt by opposing parties to deflate that image. They are part and parcel of introducing digital technologies to Argentine political culture.