In our highly polarized society, it’s not surprising that we see significant differences in how words are used by those with opposing political and cultural viewpoints. The Bridging Dictionary, an interactive web-based prototype developed by MIT’s Center for Constructive Communication, identifies the different way words and phrases are used by different constituencies and–similar to a traditional dictionary or thesaurus–gives meanings and also attempts to suggest less polarized (bridging) alternatives.

How does it work?

Utilizing natural language processing, the Bridging Dictionary compares how two media outlets on the opposite sides of the US political spectrum–foxnews.com on the right, and msnbc.com on the left–differ in the meanings assigned to the same words or phrases. This involved gathering approximately 18,000 articles from foxnews.com and 13,000 from msnbc.com published since 2021. The content was then split into millions of sentences for analysis. The first analysis measured the differences in usage frequency and sentiment. For those terms that show significant differences, a further qualitative comparison was done using a large language model (LLM) to describe the way the usage varies between the two outlets. The LLM was then prompted to provide evidence for its conclusions by citing specific references, as well as alternative "bridging" terms.

"You like potato, and I like po-TAH-to..."
For more thoughts on the Bridging Dictionary, its current challenges, and possible future potential, we invite you to visit CCC Senior Advisor and former CBS News President Andrew Heyward’s post on the CCC website.

Suggestions or feedback on how to make this better? We’d love to hear from you.

Provide feedback via the form here.

Informed by years of research focused on social media and media analytics, the MIT Center for Constructive Communication (CCC) works to combine the ancient wisdom of human conversation with emerging digital technologies to promote communication based on shared understanding and trust rather than on the binary thinking that too often divides us. Based at the MIT Media Lab, CCC leverages data-driven analytics, AI, and large language models (LLMs) to both better understand current social and mass media ecosystems and to design new methodologies and communication networks to foster more trusted and constructive dialogue. To achieve this, it brings together researchers in AI, computational social science, digital interactive design, and learning technologies with software engineers, journalists, political scientists, designers, and community organizers. An important aspect of CCC's mission is its commitment to develop improved communication channels so essential for democracies, communities, and institutions to function effectively.