Are you receiving the most up-to-date KM news? Subscribe to one or more of our newsletters to make sure you are!

Real-time digital behavior tracking—The political impact of social media

This article appears in the issue June 2017 [Volume 26, Issue 6]


   Bookmark and Share

A review of the past three presidential elections indicates the power of social media in contemporary American politics. Abundant social media channels function as among the most consistently revealing sources of big data analytics and are the vehicle through which the political—and democratic—process manifests in the country today.

However, the expansion of social media and its requisite sentiment analysis took on new dimensions during the 2012 election, in which the incumbent president’s campaign paved the way for today’s political analytics by contextualizing sentiment data with a richer array of big data sources. The result was the capability to produce real-time, digital behavior tracking, with social media acting as the means of targeting tailored advertising for constituents and potential constituents alike.

The paradigm emerging from the 2012 election was relatively simple and stemmed from the extensive social media sentiment analysis that influenced the 2008 election. Social media sites serve as the basis for candidates to detect which issues are most meaningful to voters. Political campaigners aggregate that data with other big data sources, leverage predictive models and determine how to use the information gleaned to most effectively sway voters.

The methodology requires numerous technologies, facets of infrastructure and options for analytics—many of which depend on the particular social media channel in use. A deconstruction of those various data-driven requisites sheds considerable insight into how social media, and data itself, is affecting politics today.

Sentiment analysis

The most immediate way in which social media drives contemporary American political campaigns is as a measure of sociological sentiment. Specifically, the numerous social media platforms influential as early as the 2008 presidential election yielded insight into the thoughts and feelings of the general population and of certain segments of the population (such as particular states, targeted demographics or special-interest groups). At that point in the evolution of social media’s impact on politics, it primarily operated as a medium for the general population to influence one another regarding the most pressing issues of the time. Big data analytics were in their incipient stages, so many candidates were decidedly limited in their ability to use those channels as a way of spurring action. Instead, they predominantly observed the plethora of social media outlets alongside most voters and potential voters.

Prior to the advent of big data analytics, reaping insight from social media was principally based on sentiment analysis. Such analytics were immensely affected by the very mechanisms of social media that indicated which subjects resonated with the populace and how. Prior to the present decade, those mechanisms mainly included:

  • Hashtags—Hashtags are applied to topics of interest and serve as a means of delineating which topics function as categories for discussion, commentary and other forms of social media interaction. According to Facebook’s Help Center: “Hashtags turn topics and phrases into clickable links in your posts on your personal Timeline or Page. This helps people find posts about topics they’re interested in.” Hashtags provide that same functionality on other social media sites, acting as the initial portal for subjects of note by providing links to keywords.
  • Trending topics—Trending topics are virtually identical in purpose to hashtags. They designate the subjects garnering the most attention now, which in turn fuels additional attention to create a compounding effect for further interest in those topics. Whereas hashtags are determined by users at will, trending topics found on both social media sites and others are based on analysis of demonstrable user interest in a subject or notable theme.
  • Followers—Quantifying followers is one of the most influential means by which social media channels can assert sway over potential voters. Measurements of the number of users linked to a user’s social media profile supply quantitative data of a user’s preeminence—which translates into social media power for whatever topics or stances that user advances via his or her platform.
  • Sharing—Sharing is a frequent means by which social media users can add to the quantity and degree of the effect of a posting. Nearly all social media outlets have some means by which users can replicate or link to a post, article, video or picture, which perpetuates the impact of that item. Quantifying the number of shares such a posting has is a readily identifiable means of ascertaining which subjects are resonating with users and hints at the reasons why.

The AI effect

Initial attempts at sentiment analysis for social media were purely quantitative, requiring simple aggregations of data for numbers of shares, followers and basic statistics about the impact of those data per subject and popularity of a specific outlet. Contemporary sentiment analytics of social media is much more complex, involving not only dedicated social media outlets but also an assortment of chatrooms, forums and other means by which online users communicate about socio-political ideas. Today, the majority of sentiment analysis is facilitated by various applications of artificial intelligence, which offer enhanced elucidations of an increasingly complicated array of video, image and text analytics. Technologies such as deep learning and neural networks are vital to achieving that type of analytic insight of unstructured data, partly because of their value with non-linear pattern detection and the exponential number of variables they can account for while identifying such patterns.

Modern data scientists—in addition to users savvy enough to leverage the multitude of on-demand self-service options for AI—can utilize deep learning to decipher the meaning of a variety of image and video data containing pivotal social media sentiment, which would have eluded earlier analytic methods. Natural language processing plays a crucial role in deconstructing sentiment analysis for the unstructured tweets, comments and online discussions about local and national issues that affect voting. It can ascertain meaning from slang, improper English, different languages and other factors relevant to natural language, enabling political campaign efforts to understand emergent themes in any variety of unstructured text analytics.

Contextualizing social media analytics with big data

The 2012 election remains a critical juncture in social media’s impact on politics because it represents the moment when unfolding big data analytics effectively merged with social media’s sentiment analysis, resulting in targeted campaign efforts based on digital behavior tracking. Prior to that election, social media was merely a modern way for the masses to communicate with one another, with those in political circles eagerly watching from the proverbial outside. During the 2012 election, that paradigm shifted and, courtesy of big data analytics and a newfound capability for their predictive prowess, enabled political factions to influence voters based on social media and other big data sources. The effectiveness of the latter in that regard should not be underestimated; big data analytics can include just about any particulars of life, from shopping to financial habits, entertainment choices to educational background and recreational activities.

The influx of big data largely coincided with that of data science, which was responsible for parsing through the former to identify points of relevance for an identifiable purpose. Thus, social media sentiment became merely another source to integrate and aggregate alongside other big data, which helped to form a composite of the issues of interest for potential voters. In that way, social media analytics was perhaps a fertile starting ground from which to contextualize additional big data analytics, because it served to outline specific subjects and the most relevant partisan points about them. Still, it was their integration with additional big data sources that created the digital footprint vital to the segmentation of potential voters into discernible categories with which to target campaign information. The degree of specificity involved in such segmentation (as well as the synthesis of sentiment analysis and big data) was crucial to personalizing campaign contribution funds in emails, for example, while affecting television and browser-based advertising as well.

Of equal importance were the consequences of the predictive modeling that operated at the core of the predictive analytics, which drove much of the data science exploited for political advantage. Today, that modeling process is greatly responsive to AI’s deep learning and machine learning algorithms, which can dynamically create the models derived from the data themselves at scale in time frames quick enough to service a predefined objective. Traditionally, data scientists were not only employed to successfully integrate what amounted to myriad sources, structures and types of data, but also to model them in such a way that the results of current and historic data shaped the results of future data. That process enabled the victorious presidential election campaign in 2012 in particular to determine the best way to reach voters and maximize its influence on them.

Of note in the sheer variety of big data involved in contemporary campaign efforts—which includes the many forms of social media outlets and their varying text, image and video data—are the infrastructure requirements to effectively coalesce all that disparate data for uniform analytic insight. The viability of the cloud as a medium to accommodate the scale and storage requirements of big data was invaluable, as was the ascendance of the data lake tenet as a singular means of storing all data in their native formats. Evolving semantic models (ontologies) are useful in linking such data in graph formats, which hone in on the relationships between nodes for maximum awareness of how even disparate data interrelate and are useful for a fine-tuned, campaign-specific purpose. All of those means enabled campaign efforts to forsake the traditional silo approaches characteristic of relational options, which was instrumental in aggregating sentiment analysis and big data to create segmented population predictive models predicated on digital behavior tracking.

Timely immediacy

One of the cardinal points of interest in the acute effectiveness of social media within the political sphere is the haste in which it can both deliver information and actionable analytics. The most recent presidential election typified that aspect of those technologies. The Trump campaign was notable both for its pervasive embrace of social media platforms such as Twitter as a means of fostering awareness in the public consciousness and its funding for outsourcing big data analytics. Still, the real-time potential of social media was made clear during the most pivotal moments of the campaign on Election Day. The president’s continued patronage of Twitter proved prescient when one examines the way it influenced the political landscape on election day. With prominent political action committees (PACs) and special interest groups using that medium to advertise in critical swing states, The New York Times reported approximately 30,000 posts per minute about the election transpired across Twitter, which also embedded posts into conventional newspaper outlets and popular entertainment websites, as well as featured live video streaming.

As merely one of the many social media channels used to directly convey information to the public pertaining to the presidential election, Twitter represents an indisputable example of the immediacy with which such media can influence the population during political events of enormous significance. One may also argue that social media’s ability to simultaneously transmit information and provision a rich source of analytics exceeds that of any other medium currently available, which is why it’s so vital to the modern American political landscape. It provides a direct conduit to the most meaningful aspects of today’s zeitgeist, which explains its ongoing relevance to political conversation in the country today.


Search KMWorld

Connect