“Crooked media”. We’ve heard this phrase quite often in the past year and a half and ever since then have been questioning the moral ethics of journalists working at the most popular media outlets in the world.
So what does it mean when one names the media “crooked”? Well, media bias occurs when the media emphasizes one particular point of view in a way that violates the quality of professional journalism. Some may say the press is biased for the same reason readers and the public are because the press is an establishment of people, regular people who bring their opinions along with them to everything they do.
Very frequently we hear politicians complaining about the coverage they receive from the media, but never have we seen a politician turn graphic (eg: Trump ‘fighting’ CNN). So we’ve decided to put out the fire and research the question. Is the media really biased against the current President? We wish to base research on objective tools related to the field of mathematics, probability and computer science rather than rely on human experts, as all humans tend to be non-objective and judge according to their own political views.
Luckily enough, the field of NLP (Natural Language Processing) has incredibly advanced in recent decades and has proven to be very reliable on aspects such as general sentiment analysis. In other words, it is possible, based on science alone, to detect the view of the author of an article towards the individual discussed in a specific article. We installed Stanford University NLP and applied it onto a collection of articles published by CNN (as an example), where “Donald Trump” appeared in the title of the article. We collected articles from the past 30 days and asked the Stanford NLP classifier to detect sentiments of each of these articles with regards to the entity “Donald trump”. The Stanford NLP classifier returns the sentiment as one of 3 labels: “positive”, “negative” or “neutral”. We then applied a numerical percentage value to each of these labels: “positive” = 100%, “neutral” = 50% and “negative” = 0%.
Having completed this process, we then reapplied the entire program onto the 12 most popular media websites according to their popularity. The results (July 11, 2017) were very conclusive as can be seen below.
The media is indeed much more negative towards Donald Trump than the overall public is. In fact, the media is 35% more negative about him than the votes of the free public. Truthfully, these results really were surprising with regards to the gap between how the public views the President and how the media covers him, however, the even more surprising result is that CNN (in which President Trump stated they made false claims against him) is far from being the most negative, yet in fact all of the other media sources are far more biased.
We then wanted to answer a different question. What if the media is ALWAYS negative? What if Donald Trump is not the only person that experiences negative coverage? Perhaps this is the case with everyone the media chooses to write about. After all, the media is supposed to criticize the government, politicians and public figures. We ran several tests and collected arbitrary articles about anything to make a sentiment analysis towards the entities included in each article. The answer was positive, the media is indeed more negative than positive, but the average sentiment score we received from 12 sources was ranged from 40% to 46% with very little deviation between those sources. We refactored the results to reflect on the deviation from the average source sentiment.
So yes, newspapers and journalists in general are indeed more negative than positive and the factor correction we make adds around 7% to the final scores, but still with regards to Donald Trump they display a much more negative sentiment than their average.