PoliNote

Technology article home page
article analysis from user input
political bias and accuracy score indicator
article annotations

Inspiration

We were inspired by how difficult it is to tell whether online information is truly objective or if it's hiding information. There's abundance of news sources presenting the same story differently, but honestly it can be hard for certain readers to recognize any political bias and favoritism. We live in a time period where information is often presented to reinforce existing beliefs rather than to show the full picture. We wanted to create a tool that helps visualize multiple perspectives, allowing people to see both sides of a story without bias or overlooked facts.

What it does

PoliNote is an Ai-powered political bias detector that analyzes articles for both accuracy and bias. People can search for articles or past a link for a article they are certainly reading, and the platform evaluates the content on a 0-100 accuracy scale while describing whether it's leaning more left, right, or remains neutral. People often focus on different areas depending on their interests, so there's features areas where user could read from.

This tool can also highlight specific words and phrases within the article that could indicate bias, providing reasons and annotations so the user can understand why certain languages may be misleading. Combing the scores, labeling, and in-text indications, PoliNote helps users see multiple perspectives and make more informed decisions about the information they consume.

How we built it

We built PoliNote using React and TypeScript for the frontend to create a responsive and intuitive user interface. The application should allow users to search for articles or input links, which are processed and displayed dynamically.

We used amazon bedrock for ai integration, more specifically we used amazon nova lite model to analyze article content. The AI evaluates bias based on language patterns and known information about sources, while also comparing artilces to assess factual accuracy. We wanted to design a consistent polictical spectrum scale (left, neutral, right) and implement real-time highlighting and annotation features to make the analysis more visually understanding for the user.

Challenges we ran into

One of the biggest challenges was dealing with the subjectivity of political bias. It's difficult to construct a system that evaluates bias without introducing bias itself, so we had to carefully design a logic based structure for the AI to keep it from swinging a certain side. Another problem we had was displaying the annotated text clearly. Highlighting biased phrases while keeping the article readable required thoughtful UI design and structuring. We needed to establish consistency in the system ensuring that the user was given accurate scoring within the article and not displaying its own bias.

Accomplishments that we're proud of

We are very proud of successfully building a tool that could detect bias and explains it in a transparent fashion, without displaying it's own bias. We were able to create a consistent and intuitive interface that people could easily understand even when presenting complex information, os they can quickly understand. Integrating AI into a real-time application and making it feel seamless is something we felt like wouldn't be accomplish for this project.

What we learned

Throughout most of this project, we learned how to integrated AI into a full-stack application and the importance of designing outputs that were consistently accurate. We understand that it's important to have a rubric for the AI to follow so it wouldn't hold bias for certain articles. Additionally, we gained more experience in balancing technical functionality with the user experience.

What's next for PoliNote

For the future, we want to improve PoliNote by incorporating more data sources to strengthen accuracy scoring and provide deeper comparisons across more articles. We also plan to enhance the AI's ability to detect more subtle forms of bias, such as framing and emotional tone. Other potential add-ons could be source credibility ratings and even real-time updates on breaking news for the future.