We remain a polarized nation, wounded by a traumatic election season. In fact, this polarization runs so deep, a new study at Carnegie Mellon finds that even words we use are now polarized. Viewers of different news channels are in a sense speaking different languages.
A “protest” to one set of viewers is a “riot” to another. For one, it’s a “mask,” to another, a “muzzle.”
Based on millions of user comments on the YouTube channels for four leading cable news outlets, it seems that viewers of right-wing outlets think of “Burisma,” in the same way that their left-wing counterparts think of “Kushner.”
“A very shocking example is ‘KKK’ in CNN English and ‘BLM’ (short for Black Lives Matter) in Fox News English are used in very similar contexts,” says Ashique KhudaBukhsh, a project scientist in the School of Computer Science’s Language Technologies Institute at CMU.
Similarly, it’s Black Lives Matter on CNN but All Lives Matter on Fox.
The study used machine translation tools to analyze millions of comments on the major cable news networks’ YouTube channels: MSNBC (liberal), CNN (liberal), Fox News (conservative) and One America News Network (more conservative). Everyone is speaking English, of course, but a computer analysis shows that viewers of different news channels perceive certain words very differently.
They analyzed 86.5 million comments by 6.5 million users of over 200,000 cable news videos going back to 2014. The translation algorithms uncovered patterns of “misaligned words” — terms that aren’t identical but are used in the same contexts.
“We think our method is powerful because it’s efficient,” says KhudaBukhsh. “You don’t have to read millions of comments. But if you know that ‘mask’ translates into ‘muzzle,’ you immediately know a debate is going on surrounding freedom of speech and mask use.”
That particular example comes from polarization even within conservative media, between conservative Fox News and the even more conservative One America News Network (OANN).
“Mask’ in Fox News English translates into ‘muzzle’ in OANN English,” says KhudaBukhsh. “There’s a difference between attitudes towards wearing masks during this pandemic — it’s also quite extreme.”
The project started, ironically, by something KhudaBukhsh was studying during escalating tensions between India and Pakistan.
“I previously worked on ‘hope speech,’ with two countries on the brink of war,” says KhudaBukhsh. “We were trying to use artificial intelligence to find content where people were talking about peace, about hostility-defusing. So if we can find voices that bring the country together, bring the two polarized communities together, that would be really something useful to do.”
That gave them the idea to test the differences between the discourse on CNN and Fox. Using an automated translation algorithm, most of the words would be the same — they’re both speaking English — but some of them are very different.
“If they’re different — which means these two words are used in identical contexts, but the words are different — that might tell us a lot about the two communities.”
Perhaps unsurprisingly, these misaligned pairs extend to the point of insult.
“‘Trumpty’ in CNN English translates to ‘Obummer’ in Fox News English,” says KhudaBukhsh.
But there are also clear demarcations when it comes to actual policy disagreements.
“Solar’ in CNN English translates into ‘fossil’ in Fox News English,” says KhudaBukhsh. “This actually tells us a very interesting story about where our differences are coming from. We hear something like solar and fossil, we definitely know there’s a debate surrounding the energy sources. The moment we hear something like KKK/BLM, we know there is a debate surrounding racial injustice.”
In addition to finding misaligned pairs, they calculated the differences between the languages of MSNBC, CNN, Fox News and OANN. Words translated from MSNBC English to CNN English had 63% similarity, while words translated from MSNBC English to OANN English had only a 42% similarity.
They even decided to analyze late-night comedians Trevor Noah, Seth Meyers, Stephen Colbert, Jimmy Kimmel and John Oliver. They found words translated from Fox News English to ‘comedian English’ were 75% similar, while words translated from CNN English to ‘comedian English’ were 83% similar.
So the question remains — will this polarization continue? Will our words themselves continue to drive us apart?
“I don’t know,’ says KhudaBukhsh. “One thing that I can tell you that we found reassuring was that in all these news languages, ‘democracy’ always translated to ‘democracy.’ So we are on the same page for many of the concepts and words, and democracy is the most important.”
The authors of the study include Rupak Sarkar, Mark Kamlet and Tom Mitchell, all from CMU. A preprint of their paper, which has been submitted to a computer science conference, can be read here.