Inspiration

The amount of political bias present in current forms of news reporting and social media can easily trap unsuspecting readers into echo chambers constructed of misrepresentations of the truth.

What it does

This AI-powered Chrome extension identifies the bias present in a tweet and provides a neutral, unbiased summary of the topic while also representing opposing viewpoints in an user interfaced embedded within Twitter itself.

How we built it

We created a vanilla extract-based javascript extension, utilizing web scraping to pull the text of a twitter post that the current tab is on. It then takes the data and feeds it into Google's Gemini AI model via its Rest API. The large language model returns a response including all the aspects described above and embeds it into the Twitter web page right below the tweet in a clearly visible manner so the reader can eliminate any personal biases after evaluating the tweet and taking in account all newly presented information.

Challenges we ran into

One of the biggest challenges was learning how to use API's, specifically working around Twitter's never-ending restrictions their API as a free account. Additionally, evaluating different frameworks (like React.js) in conjunction with the Chrome Extension and learning how to effectively use the Large Language Model on both text and attached media proved challenging. At the end, we went with vanilla javascript without any frameworks because it proved to include too many unnecessary complexities to be a reasonable aspect of our 24 hour project.

Accomplishments that we're proud of

We learned to make a feasible chrome web extension that could be adopted widely and have potential considerable impact on society.

What we learned

We learned how to use APIs and the Chrome extension workflow.

What's next for Bias Identification Service

Support for images and other social media sites besides Twitter. Also account support via Auth0 to keep hard track of tweets (documents it even if it get deleted).

Built With

Share this project:

Updates