The divide between political parties is not just ideological. Their members are essentially speaking different languages, a study from researchers at Carnegie Mellon University found.

Using computer translation tools, researchers analyzed 86.6 million YouTube comments from 6.5 million users on more than 200,000 news videos from CNN, Fox News, MSNBC and One America News Network (OANN). The dataset included comments ranging from 2014 to July 2020.

“We wanted to analyze: If we consider one news viewership and the way they speak — the way they write — how is that different from another news viewership, the way they use the English language,” said Ashique KhudaBukhsh, a project scientist in the School of Computer Science’s Language Technologies Institute and one of the study’s co-authors.

They found that viewers of right- and left-leaning outlets often interpreted the same words in vastly different ways — almost like they were using different words altogether.

The researchers used machine translation algorithms that are typically used to translate from one language to another — English to Spanish, for example. They found that some of the same words translated differently, depending on the contexts in which they were used.

It’s similar to comparing American English to British English, KhudaBukhsh said. They are mostly understandable between each other, but some words differ when talking about the same things. “An apartment” in America is “a flat” in Britain, for instance.

The study found that different language was used to describe the same things or people, and often, the results were striking. “Mask” for some left-wing viewers translated to “muzzle” for some right-wing viewers. “Black Lives Matter” to some right-wing news viewers was often used in the same context as “Ku Klux Klan” for left-wing viewers.

Comparing the language from one news network to each other, the researchers found that left-leaning MSNBC and far-right OANN showed the most contrast, with just a 42% similarity in words and contexts.

“It’s OK if people have very different opinions. That’s not necessarily highly-alarming for the health of a political system,” said Mark Kamlet, a professor of economics and public policy and another co-author of the study. “But when we get to the point where there’s so little in common, there’s no basis in having an intelligent conversation.”

The study is still being peer-reviewed, pending potential publication and presentation at an artificial intelligence conference. But the researchers said they decided to share their findings ahead of the election. The team includes Tom Mitchell, Founders University Professor; and Rupak Sarkar, research engineer for a fall 2020 seminar course on tracking political sentiments using machine learning taught by KhudaBukhsh, Kamlet and Mitchell.

The study seems to raise flags about the way language is used in different subsets of communities, the researchers said.

“Civil discourse,” KhudaBukhsh said, is “dissipating.”

He considered his experiences as a child, growing up in India and viewing TV news as facts with no opinions attached. Now, with viewership so polarized, he said the results raise questions about whether that’s still the case for some networks, and if there is some degree of misinformation within certain groups. He also noted that both left- and right-wing viewers had a tendency to stoop to name-calling and insults.

Political disagreements are normal, Kamlet said, and some degree of extreme partisanship is not a uniquely American experience. But when people with differing views aren’t speaking the same language — and don’t seem to have the same fundamental understanding of the facts — Kamlet said there’s evidence of severe polarization.

“In some ways we wish we didn’t find what we did,” Kamlet said. “When you don’t have enough in common, there’s nothing for you to discuss with one another. And when a society, when a political system gets in that shape … there certainly isn’t a real easy game plan.”

Kamlet said he hopes the study functions as a wake-up call — an opportunity to improve communication. The co-authors said Carnegie Mellon now has an entire course dedicated to studying the dataset and looking for different language patterns.

More optimistically, KhudaBukhsh said there were still some commonalities among the viewership.

“In all news languages, ‘democracy’ still translates to ‘democracy.’ We are still on the same page about what democracy means to us,” he said. “I think that’s a good thing.”

Copyright ©2025— Trib Total Media, LLC (TribLIVE.com)