Browse By

Looking at Polarization Through Machine Translation

Written by Gloria Wang

It is not news that US politics has become increasingly polarized. Yet, the chaotic presidential debate that occurred earlier this year came as a surprise to many Americans. In a recent study, researchers at Carnegie Mellon University have discovered that the debate was only one of the symptoms of the nation’s collapsing civil discourse, a result of the growing disparity in the polarization of words. Although everyone is speaking English, machine translation analysis shows that viewers of news channels with different political biases are, in a sense, speaking different languages.

Modern machine translation methods determine the meaning of a word based mainly on its context. While this method is traditionally used to compare words in different languages, the team at Carnegie Mellon used the same method to analyze the polarization of the English language in America. For instance, a conservative might say “Democrats are the greatest threat to America today,” while liberals might say “Republicans are the greatest threat to America today.” Democrats and Republicans are used in the same context, making them misaligned pairs. Although this is a very simple yet extreme example, it demonstrates a common usage of political polarization (Spice 2020). 

The study uses a data set that consists of over 85 million comments on over 200,000 videos from 6.5 million viewers since 2014, across a multitude of Youtube videos from MSNBC, CNN, Fox News, and One America News Network (KhudaBukhsh 2020). With this data set, researchers analyzed the variations in language use in the comments section, similar to the distinction between British English and American English. Using machine learning methods, it was discovered that these differences do exist. This finding serves as the first demonstration of quantifiable linguistic differences in news audiences. 

Yet perhaps even more surprising, as the team explains, “the system found that words with vastly different meanings, like ‘KKK’ and ‘BLM,’ were used in the exact same contexts depending on the YouTube channel being analyzed” (Kamlet 2020). In other words, the comments made by one community about the KKK, the Ku Klux Klan, are very much like the comments made by the other about BLM, Black Lives Matter. While the beliefs of the KKK and BLM are about as different as can be, depending on the comment section, they seem to each represent something similarly ominous and threatening regarding opposing political parties (Spice 2020). 

Besides having divisive, and heavily polarized political views, this study shows that even the language that all Americans are supposed to share and communicate through is divided in and of itself. While having a variety of different opinions within a country is actually beneficial to a democracy, it is crucial to understand the necessity for civil discourse in order to maintain a functional nation of civil discourse. Only through patience and the willingness to listen to each other can American English become one “language” again.



KhudaBukhsh, A., et al. (2020). We Don’t Speak the Same Language: Interpreting Polarization Through Machine Translation. arXiv.  arXiv:2010.02339v2. Retrieved 22 November 2020, from 

Spice, B. (2020). Even Our Language Is Polarized. (2020). CMU. Retrieved 22 November 2020, from  

Kamlet, M., et al. (2020). Fox News viewers write about ‘BLM’ the same way CNN viewers write about ‘KKK’. Retrieved 22 November 2020, from  

Leave a Reply

Your email address will not be published. Required fields are marked *