DemocratAIze

Inspiration

Our inspiration for creating DemocratAIze is to overcome the cloud of political noise and rhetoric entrenched in modern politics, especially in North Carolina, a contested battleground state. Whether it’s yesterday’s flashy headlines or today’s misappropriated quotes, the sheer abundance of political rhetoric can make it hard to gauge where exactly a political figure stands on key voter issues.

This is where DemocratAIze comes in, offering a personalized political research tool for NC politics that allows you to understand your civic environment, where candidates stand, and how much your own political beliefs align with theirs.

What it does

The purpose of DemocratAIze is to help voters understand the candidates in their districts and see just how much they share the same political beliefs, from reducing the cost of living to regulating AI. First, users answer a questionnaire, inputting their political beliefs on pressing issues. Then, the user ranks their top political issues including cost of living, healthcare, immigration, etc, before putting in their zipcode. Based on that zipcode, we pulled local politician data and quantified their perspectives and talking points, then compared it to the user's political beliefs. From there, users can read more about where that candidate stands on the aisle, with information on their political promises.

How we built it

We utilized a Python script to extract NC political candidate information from the US Census GeoCoder API, including the candidate’s name, photo, affiliation, and geographic boundary of their district. Then, Gemini Flash 3.8 was given web search tool use capability to find detailed information on each candidate from a diverse set of predominantly primary sources, including the candidate’s website platform, official social media posts, transcripts of speeches or interviews, and legislation they introduced, if applicable.

Using this information, Flash 3.8 then evaluated each candidate on the same scale of -2 to 2 that our users were placed into based on their selections. The candidate information, district information, Gemini’s research on their policies, and its overall numerical score were stored in four respective tables in TigerData. Based on the user’s provided ZIP code, the candidates running in their various districts were identified, and the squared difference between the user and each candidate’s numerical score was calculated. Then, a weighted average was taken of the squared differences, based on the weight given to each political issue in the user’s rankings.

Based on the weighted average, a percent score is displayed showing the user’s match with each candidate, and the user has the option to view a dropdown menu showing where they stand relative to the candidate on each key issue. The user can also better understand the overlap between their views and those of the candidate through AI summaries powered by Gemini.

Challenges we ran into

Figuring out how to properly present a politician’s credentials and beliefs without appearing biased. A measure we took to resolve this challenge was to divide a candidate’s available information into two categories: what they’ve done and what they plan to do. This way, users don’t just have to rely on a candidate’s claims and promises to see if their beliefs align with the users'. Instead, the voter can directly see what their candidates have done in the past.
Properly gauge voters' political beliefs. We found it was difficult to represent the wide and diverse range of voters’ political views on topics such as the economy and foreign policy. Hence, we decided to implement an option for users to manually input their own beliefs to be scaled.
Dealing with candidate data that lacked relevant or plentiful information. For these specific races, we took an honest approach and decided to let the user know that these races lacked sufficient information.

Accomplishments that we're proud of

Since the midterm elections are less than a month away, we are proud of creating a product that can create real-world impact by helping users find candidates that reflect their values and helping them connect with those candidates. Most of our members are first-time hackathon participants, so we are proud of working together to create a compelling product in a relatively short time frame, despite the brainstorming challenges, countless caffeine drinks, and post-midnight grinds.

What we learned

We learned the importance of delegating tasks based on our unique skill sets in order to effectively work together and create a compelling product. Harshil primarily worked on creating the script to collect candidate data and building the backend in NextJS. Shravan worked on creating an aesthetic and user-friendly website. Eric worked on planning the product and brainstorming features, while Alex worked on the presentation of our product.

Given this is the first hackathon for many on our team, we also learned to incorporate new tools such as Tiger Data and Lovable during the development of our app.

What's next for Democrataize

During planning and development, several ideas came up that we couldn’t implement, either due to time constraints or to prevent scope creep. However, in the case of future development, several ideas we’d implement include expanding the app to cover municipal and national elections. We find this idea especially promising for the latter not only because the increased misinformation and rhetoric that accompany national elections are the perfect target for our app. Another idea we had would be to integrate political fact-checkers into our app to verify political claims with previous actions, or to see if incumbent candidates fulfilled their previous promises. This way, users can not only analyze the political promises of candidates but also how truthful they may be about carrying out these promises.