In this blog, Senior CCC advisor and former CBS News President Andrew Heyward explores the reasoning behind the Bridging Dictionary, acknowledges some of its current challenges, and offers an invitation to provide your feedback for its future development.
You like potato, and I like po-TAH-to…
You like tomato, and I like to-MAH-to…
Potato, potahto, tomato, tomahto –
Let’s call the whole thing off.
– George and Ira Gershwin, “Let’s Call the Whole Thing Off”
What if generative AI could help us understand people with opposing views better just by showing how they use common words and phrases differently? That’s the deceptively simple-sounding idea behind a new experiment from MIT’s Center for Constructive Communication (CCC).
It’s called the Bridging Dictionary (BD), a research prototype that’s still very much a work in progress – one we hope your feedback will help us improve.
The Bridging Dictionary identifies words and phrases that both reflect and contribute to sharply divergent views in our fractured public sphere. That’s the “dictionary” part. If that’s all it did, we could just call it the “Frictionary.” But the large language model (LLM) that undergirds the BD also suggests less polarized alternatives – hence “bridging.”
In this prototype, research scientist Doug Beeferman and a team at CCC led by Maya Detwiller and Dennis Jen used thousands of transcripts and opinion articles from foxnews.com and msnbc.com as proxies for the conversation on the right and the left. You’ll see the most polarized words and phrases when you sample the BD for yourself, but you can also plug any term of your choosing into the search box. (For a more complete explanation of the methodology behind the BD, see https://bridgingdictionary.org/info/ .)
Check out some of our “favorites” – believers, radicals, domestic terrorism, misinformation, woke, republic, nationalism – but you’ll quickly find others that illustrate why we too often seem to be talking past one another rather than to one another.
You may find the word “dictionary” a bit misleading: you won’t find sharp differences in how words are defined so much as stark variations in how they are used: contextual conflict, if you will. But you will quickly see how these variations reinforce the hardened silos of our public sphere.
What do we hope to accomplish with this experiment? We believe it offers a unique lens through which to view our polarized conversations. We would love anyone who is interested in fostering healthier communication to use the Bridging Dictionary as a source of reflection and perhaps inspiration.
We can also imagine academic researchers and journalists finding value here, although we’re not sure who is likely to use the suggested alternative language. (When we showed the prototype to our friends at PBS’s FRONTLINE, they were impressed by the insights but skeptical that journalists would be eager to embrace substitute words offered by an algorithm.)
We are making the BD publicly available in this early stage of its development in the hope that users will help us realize its full potential. We would like to learn whether and how the BD might be useful.
We also welcome feedback on its limitations. The most obvious is that it is based on data only from two cable news channels, so the tool is analyzing the language of pundits, politicians and personalities rather than viewers and readers. An earlier prototype was based on (now outdated) Twitter posts. Social media would seem like the most obvious way to parse the language of large groups of users, but that data is increasingly hard to come by. Suggestions welcome!
Most importantly, Doug and the team built this research prototype in the belief that understanding divergent use of language–even in this early, imperfect prototype–can help people bridge divides and connect more effectively. We hope you will help us not only improve the Bridging Dictionary, but test this premise.
– Andrew Heyward
Senior Advisor, CCC & Former President, CBS News
