Digital Democracy: Sentiment Analysis in a Polarized Multi-Party Political Context

Internet has permeated social relations to such point that it has come to determine the ways in which society communicates and interacts. And, as such, policy and political behaviour do not escape the scope of changing technologies. From desktop computers to smartphones, connectivity and ubiquity establish an interaction structure which represents the very importance of Information and Communication Technologies (ICTs), where every user can be a producer of data and information (prosumer) with no limits to his or hers audiences. The debate, however, turns to the appropriation of such advantages within the Web; now that everyone has the means to express themselves the way anyone wants to, who is being heard? Facing this paradox, social media (SM) have come to play the gatekeeping role, leaving users in (partial) control what is seen, heard, or spoken. Politics, again, does not escape from this. Political engagement has risen to be a new topic of discussion among scholars in order to comprehend social networks in the midst of a technologically-mediated environment and to grasp the levels of influence some actors (or nodes) have among their networks. Based on such intricate paradigm, political interaction raises broad fields of study, leading to a better understanding of social phenomena with respect to electoral and non-electoral participation. However, most of the academic work today lacks a point of view that isn’t European or American-centric. This, despite the valuable results in terms of analyses and democratic studies, weakens the global aspect of ICTs, as well as the perception of particular realities which determine the value of democratic nation-states, such as economic disparity, social cohesion and digital divide. For this reason this paper will focus on the specific context of the 2018 Costa Rican presidential elections, where up to 13 candidates were running for office, displaying the level of multi-polar political spectrum in the country. The analyses will be based on the premise that the digital an “offline” worlds are deeply interconnected, and that any activity perceived online will represent, with sufficient certainty, any offline actions. To do this, the research focuses on the top eight candidates’ official Facebook pages. Using an open access tool for extracting data; these pages will be mined during the whole campaign period (October 2017 – February 2018). Then, descriptive statistics will be used in order to present common digital communication trends, unusual activity and to help describe the correlation of the virtual public sphere with the offline political environment. Additionally, a sentiment analysis (SA) will be carried out based on all the comments extracted in order to determine the polarity of each candidate’s support. This represents an additional challenge given that there are very few tools available for doing this in Spanish. The R package qdap is used in this paper to process this part. Other packages will be used to present other content analysis values, like predominant topics and word correlations. This will evidence just how discursive politics in polarized contexts (predominantly, social media) carry an important weight in people’s electoral pondering.