By Daniel Tanev Parapunov, a 4th semester DIM student at the ITU who used ANTA through ETHOS lab as part of his master thesis on the topic of how digitalization is changing the news media.
Just a month short of the exams, I realized how utterly unfinished my research was, especially in its aspect of measurability. Despite case-studying one of the biggest overseas news agencies, I was still struggling to find data, as media giants like these would not open up their empirical treasury to the public simply by request, and especially not to people who were trying to make them look bad. After all, investigating the impact of digitalization on media freedom and access to quality news, was not exactly something the Cable News Network (CNN) would want to be part of, as they were already at the heart of this degenerate practice and going strong. But setting my high moral beliefs aside, I still needed something with which to back up my project quantitatively. Fortunately, the Actor Network Text Analyzer or ANTA in short, proved useful in this situation of need, and, most importantly, had the full capabilities of turning CNN’s own weapons against them.
Digitalization has been one of the primary driving forces behind recent changes in journalism including news values, professional ethics and newsroom management. You would not need a team of scientists to tell you how with the improved access to information and its channels of distribution, plagiarism and lack of verification have severely infected mainstream media. On top of that, our contemporary digitalizing environment is fostering a need for instant gratification and loss of patience in people. The oversaturation of senses, as an effect of digitalization, is staring to alter humans’ ability to address complex challenges and to form critical opinions about the world.
A research done by Microsoft in 2015 had found that living in the digital age had made it difficult for people to stay focused, with the average human attention span alarmingly shortening from 12 seconds to 8 seconds in more than a decade[1]. If you think about it, this phenomenon offers new opportunities in the context of contemporary news media to such an extent where “it is no longer even necessary for media institutions to attempt to hide their blatant work of manipulating public opinion, manufacturing consent, and creating winners and losers in the minds of the already brain-washed public”[2].
Ultimately this increased access to technology and the consequential lowering of the people’s attention span, had resulted in an all-time high media viewership and a shamefully low journalistic accountability. The erosion of news standards, conveniently accelerated by digitalization, had made it habitual for journalists to “commit plagiarism and to pay less attention to verifying the facts and sources for their stories”[3], with some news agencies turning this into a full-time agenda.
I wanted to quantify how severe the issue of news trustworthiness really was, especially in the case of an American television channel giant and its online editorial. I set out a hypothesis that CNN structures their stories in a propagandistic manner, by implementing a strong bias right from the beginning of their news articles, and this is where ANTA came into play. Before diving deeper, however, first I had to narrow down my research scope and find a topic which was naturally susceptible to biases. Luckily enough, at that time the presidential election was at the peak of its heat, with media agencies spewing controversial stories by the minute. With the Democratic and Republican parties being the “main contenders”, my job of finding possible bias in CNN news reporting became easier. It was really straightforward to identify how frequently Hillary Clinton and Donald Trump appeared as actors in the discourse, which in this case was narrowed down to some 140 articles out of the hundreds more, composed by CNN during the full duration of the 2016 election.
Going back to my hypothesis, however, there was one additional measurement needed, which would reflect on the lowered attention span in today’s average citizen. Putting it into perspective, people nowadays would read only the beginning few sentences of an article and move on to the next one, without getting the full story. Cunningly enough, news agencies could be using this trend to manipulate the social opinion by structuring their articles with a strong bias from the very beginning.
In order to measure whether this is true I had to conduct 2 separate scans of those 140 articles through ANTA, which I labeled as 2 “News Realities” for easier differentiation:
Reality A was going to be the result of analyzing the full text corpus of the articles, and
Reality B was going to be the result of analyzing only the first 2 paragraphs of each article.
It was believed that if any viable measurements of differentiation between the two scans or realities were made, it would serve as proof for my hypothesis. The measurements themselves where made with another tool called Gephi, which I used to transform ANTA’s findings into a visible, lifelike representation of the discourse. Hold on, it is going to get a bit technical!
Figure 1 is a visual representation of the fully scanned articles’ text corpora. A network complexity became evident with numerous leading components and sub-components (nodes). The labeling of each node was automatically facilitated by the text analysis done via ANTA, by algorithmically identifying different subjects, topics and objects within the articles. The size of each note represents the level of connectivity it has to other nodes in the systems, and its frequency of appearance in the “discourse”. The unlabeled nodes represent the articles themselves as individual entities. The color of each node speaks for its modularity, where different colors represent different “communities” within the network, separated in a logical way. The scale of the arrows between different nodes, indicate the strength of that connection, with bigger arrows hinting towards a more direct and unconditional relation.
Figure 1: Actor-network visualization of CNN’s 140 articles in their full scanning, on the US presidential debates
Evidently, Donald Trump and Hillary Clinton, being the main presidential candidates, are in the center of the network, standing out with numerous connections to other entities in the system, with other actors, including CNN itself also present in the visualization. Essentially Figure 1 served to illustrate the existence of Reality A, which I spoke of earlier, where an actor-network scan of the full articles represents a complex entanglement of entities and their relationship, in the political environment of the presidential debates. Metaphorically speaking, Figure 1 is how the complete written (text) coverage of CNN on this topic would look like as an image: a reality with no considerable bias and built around 2 main actors, each centered within their surrounding modularity.
I then moved on with Figure 2, which is the visual result of analyzing only on the first 2 paragraphs of the 140 articles from CNN, with the size and color attributes of the nodes behaving in the same way as in Figure 1. However, unlike the two 2 principal entities portrayed in the previous visualization, in Figure 2 Donald Trump is identified as the sole main actor in the network, with an incredibly high amount of connectivity to all other actors. Although a distinction by modularity is also present, the sheer size of the leading node within the system indicates that a systematic bias has been applied, where the relativity of the network’s entities is solely based on their unconditional relationship with the main actor.
Figure 2: Actor-network visualization of CNN’s 140 articles in their beginning 2 paragraph-scanning, on the US presidential debates.
In more practical terms, my findings served to prove that the first 2 paragraphs of the 140 articles composed by CNN on the presidential elections, have been contextually entangled, predominantly with one of the presidential candidates, namely Donald Trump. My goal, however, was not to identify CNN’s attitude towards this actor (which I cannot do anyway), but only to illustrate the raw frequency in which Donald Trump has been addressed by the news editorial over a specific number of articles.
In the context of my research, Figure 2 illustrated the existence of a very specific type of news “reality”, where an actor-network analysis of the first 2 paragraphs of the articles point out towards one giant component with sequential, negligible in size and importance sub-components.
I was not surprised, really. Should CNN have striven for objectivity, then a full-scale representation of the real world arguably cannot differ from any alternatively sized representations. In context, this meant that Reality A, being the result of CNN’s maximum textual output, should not differ from any other “partial” realities, such as Reality B, that involve only segments from the news media’s article corpora.
I believe it is fair to mention that I completely acknowledged the fact that CNN’s journalists cannot reflect consciously on the structure of their writings, in the sense that is humanly impossible for each major segment within an article to be an objective representation of the news. However, when systematic patterns were continuously becoming present across a majority of articles, such as the predominant mentioning of one actor at the beginning of every article, that alone served as an adequate reasoning to suspect acts of directed manipulation. Deceptive objectivity is what I like to call it.
In particular, according to my analysis, CNN have purposely structured the presidential election coverage, so that their articles always begin by mentioning Donald Trump in every sub-context or situation, and then continuing on to other actors and topics within the discourse. In relation to the claim regarding the inability of the average news reader to focus for extended periods of time, I saw fit the argument that CNN are using this lowered attention span to manipulate the social opinion. How, you might ask?
Well, ultimately, people who would only read the first 2 paragraphs of an article would be facing Reality B, where the presidential elections are exclusively focused on Donald Trump. In order to get “access” to Reality A, and its seemingly unbiased and more detailed reflection on the elections, viewers would have to familiarize themselves with the full articles. The differences between these 2 realities was crucial enough to justify my hypothesis that CNN structures news stories in a propagandistic manner, by implanting a strong bias right from the beginning of their articles. And it was something which I could not have proven without the help of ANTA
One thing is for sure, though. Digitalization has facilitated a tremendously improved access to information, but at the cost of widely spreading unethical practices amongst news reporters, the most worrying of which being plagiarism, lack of verification and impertinent manipulation. To make matters worse, technology has implanted the seed of neglect in individuals, where people are progressively becoming less focused and more ignorant about their surrounding political, social and economic environment. So next time when you are waiting in line at the grocery store, think twice before jumping to conclusions, through the ignorance of your 5 inch screen of wonder.
[1] Microsoft (2015); “Attention spans. Consumer insights” Microsoft Canada, accessed 20th November 2016; https://advertising.microsoft.com/en/WWDocs/User/display/cl/researchreport/31966/en/microsoft-attention-spans-research-report.pdf
[2] Activist Post (2014); “6 Examples of Media Manipulation”, accessed 20th November 2016; http://www.activistpost.com/2014/03/6-examples-of-media-manipulation.html
[3] Open Society Foundations (2014); “Digital Journalism: Making News, Breaking News”, accessed 20th November 2016; https://www.opensocietyfoundations.org/sites/default/files/mapping-digital-media-overviews-20140828.pdf; pp. 109