Introduction

The term ‘Fake News’ appeared and spread rapidly during the 2016 Presidential election. Misleading articles and stories were being published on various outlets to serve the interests of political parties. Donald Trump continues to make claims about our “deceptive” media and dispels their content as fake. However, pushing fake news to the masses is a controversial topic for Democrats, Republicans, and the public. Fake news may play a role in both political parties in helping them achieve their desired narrative.

In order to understand the type of content certain news outlets are publishing, a dataset developed by Buzzfeed News will be analyzed. This dataset includes Facebook posts from several news outlets, the direction they lean (left, right, mainstream), the factuality of the post, and how these posts are being interacted with. The Facebook posts in this analysis range from September 18th to the 26th in 2016, prior to the election.

To understand how factual the posts are, Buzzfeed News provides the descriptions behind their ‘Truth Ratings’:

Misinformation is dangerous and can skew the perception of the world for many individuals by putting a biased lens over their eyes. Whether the information is misleading or completely non-factual, it is a major issue with our media and politicians today. With the ability to publish news and stories online, especially by major news outlets, information quality is an increasingly important issue.

The proliferation of large-scale social media data paired with it’s use as a primary news source is forcing a re-examination of accuracy and credability of the text inside these stories (Buntain and Golbeck 2017). Facebook and other outlets such as Twitter have allowed stories containing misinformation to be published without review, which then reaches the masses. Those using social media have a bias towards believing what those in their network share, which allows these stories to propagate through multiple platforms (Buntain and Golbeck 2017).

Research Questions

  1. What type of content are the different leaning news sources publishing on Facebook?
  2. By news outlet, who is publishing non-factual stories?
  3. What news categories are more likely to publish non-factual content?

What Type of Content are Different News Sources Publishing on Facebook?

When looking at the distribution of content by category, right wing news sources publish the highest amount of mostly false and a mix of true and false content to Facebook. Mainstream media appears to publish almost entirely factual news which would be expected as they are informing most of our country. Left wing outlets publishing to Facebook also publish mostly false and a mix of true and false news, but not as high as the right.

25% of all content left wing sources publish fall under no factual content, which include opinion and satire posts. This may be a tactic used by left wing sources to hide bias in their content while still trying to push a message through. This is performed by the mainstream and right media, but the highest percentage falls under the left. Despite a lot of factual content posted, the left and right are still posting deceptive content which will misinform the public.

By Outlet, Who is Publishing Non-factual Stories?

Breaking down the content published by the specific outlet by category provides a lot of information. Starting with the left, almost 25% of what Occupy Democrats and Addicting published to Facebook is a mixture of true and false content. For The Other 98% and Occupy Democrats, more than 25% of what they published contains no factual content.

The mainstream media pubslishes almost entirely factual content to Facebook, with a minimal amount being mixed. There is also a small amount of no factual content posted.

When analyzing the distribution of content from the right wing sources, there is a large amount of mixed content posted. More than 25% of all of Right Wing News’s posts are a mixture of true and false content. Freedom Daily only posts factual content 50% of the time. The other 50% is almost entirely made up of a split of mostly false and mixed truth content. A quarter of Eagle Rising’s posts contain no factual content, with under half of all published content being mostly true. Around 20% of their posts are a mixture of true and false information.

Linear Model

Statistical models
Model 1
(Intercept) 18025.28***
(1381.60)
Categorymainstream -17864.62***
(1631.91)
Categoryright -16775.37***
(1787.18)
R2 0.05
Adj. R2 0.05
Num. obs. 2212
RMSE 29013.64
p < 0.001, p < 0.01, p < 0.05

A linear regression is performed to understand how content is shared on Facebook depending on the category. Content from the left is shared ~18,000 more times on average than content from mainstream media, and ~17,000 more times on average than content from the right.

Logistic Models

Next a logistic regression is performed to obtain the probabilities of fake content being produced depending on the category. The mainstream media is excluded as they have shown to publish content that is mostly factual. Left wing news sources have a simulated probability of 5%, whereas the right has a probability of 12%.

##        V1         
##  Min.   :0.02256  
##  1st Qu.:0.06463  
##  Median :0.07565  
##  Mean   :0.07547  
##  3rd Qu.:0.08606  
##  Max.   :0.13723

The results of this regression show that right wing news sources have an 8% higher probability of publishing mostly false content to Facebook than the left.

A second logistic regression is performed on the left and right, but only this time ‘mixture of true and false’ content is included. The simulated probability for the left to produce mostly false and mixed true and false content is about 19%, whereas the right comes in at ~38%.

##        V1        
##  Min.   :0.1005  
##  1st Qu.:0.1660  
##  Median :0.1838  
##  Mean   :0.1849  
##  3rd Qu.:0.2027  
##  Max.   :0.2886

The results of this regression show that right wing news sources have a 19% higher probability of publishing mostly false and mixed true and false content than the right.

Conclusion

The results from this analysis conclude that both the left and right wing news outlets publish mostly false and a mix of true and false content. However, right wing news outlets publish much more false content and much more of a mix of false and true content to Facebook. For the three right wing news sources, almost 50% of their content is a combination of mostly false and mixed content. The highest percentage of this combination for the left comes in at 25% for only one source, Addicting Info.

When looking at the probabilities of publishing fake content, right wing news sources have higher chances of sharing this type of content to Facebook than the left. When mixed true and false content is included in the model, the probability of right wing news sources sharing this content goes from 12% to ~38. Left wing news sources still have a relatively high probability at 19%. The inclusion of mixed true and false content for left wing news sources increases the probability from 5% to 19%.

It was interesting to see that left wing content on Facebook is shared more on average than right wing and mainstream media sources. Even though right wing content is shared less than the left, there is a much larger amount of misinformation within what they publish.

Overall, misinformation in social media is a problem that both left and right wing news sources contribute to. Despite the right publishing much more misleading information than the left, both sides engage in this practice and we as the public suffer. Publishing misinformation and deceptive content to further certain agendas is damaging to our country and our democracy.

References

  1. Fact-Checking Facebook Politics Pages, 2016. [Online]. Available: https://github.com/BuzzFeedNews/2016-10-facebook-fact-check, December 2017

  2. C. Buntain and J. Golbeck I Want to Believe: Journalists and Crowdsourced Accuracy Assessments in Twitter. arXiv:1705.01613v1 [cs.SI], December 2017