PoliticsForYouth

Inspiration

Throughout history, we’ve seen how politics can impact every facet of our lives. Just the recent debates surrounding the Department of Education budget cuts show us how a single decision can cost hundreds of millions of dollars and affect people across the country.

During the 2024 presidential elections, we were surprised by the lack of political awareness among our peers. After sending out surveys, we realized 75% of students were interested in politics but did not pursue their interest because of a lack of resources.

Schools avoid discussions around political topics, and communities often expose students to a single ideology. In addition, news is often biased and requires prior knowledge, making it unappealing to youth. The low political awareness among youth means a higher chance of continued low engagement as they begin to vote. This leads to uninformed voters, low voter turnout, and political polarization, among various other problems.

An issue rooted in the dullness of existing sources demands a creative solution, inspiring our president simulation game!

What it does

PoliticsForYouth is a game that guides users through the process of making decisions that affect a variety of sectors of society. It introduces players to the types of actions taken by the US executive government and points them to the impacts of actions based on historical trends. Realistic scenarios with uncertainty and variability keep the game interesting even after hundreds of plays! It's not unlikely to be presented with an option you'll never see again.

The player starts by choosing one of five actions that have been featured in the media in the past year. From there, we predict the potential impacts and use the severity of those impacts to decrease or increase a player's economy, war risk, and social welfare scores. In order to win the game, a player has to maintain a an economy and social welfare score above 90 and a war risk score below 10 for three turns, navigating risky decisions and surprise events (like a pandemic). If a player ever escalates to war (war score above 90) or their economy or social welfare scores drop below 10, the player automatically loses. Each game promises valuable context for current events and a different spin from any previous rounds.

How we built it

We started by generating potential actions based on the past year's political discourse, using gpt-3.5-turbo, to seed the first five actions for a given game. From there, we generated lists of potential impacts, based on historical trends and the action's categories (like how tariffs are going to disproportionately effect the economy), and calculated probabilities of different impacts occurring to increase the variability in which impacts would occur. We combined these into an algorithm that repeatedly generates a logically and politically relevant set of impacts following from each decision the user makes.

Our main goal was to increase relatability and realism, so we set up random events (like climate change, and pandemics) that change player's scores at varying times, simulating unforeseen events that occur in real life. To improve accuracy, we trained a classifier model on a few datasets to more accurately predict political leanings of actions, ensure the actions were based on current discourse, and influence the ensuing impacts from a given option.

Lastly, we used React.js for the UI, integrating it with Python and our classification model to run our code.

Challenges we ran into

The main challenge we ran into was how different this project was from any of the other projects any of us have done. This was the first time we created a game or a website -- we're mostly used to coding problems and task-oriented projects -- so there was quite a lot of trial and error that went into logistics for the game and planning out the best event flow.

We pretty much had to learn React.js on the fly since none of us had much experience with web development!

Accomplishments that we're proud of

We're proud of how much we were able to achieve with relatively little experience and time! The fact that the generations are so accurate and constrained to prevent LLM hallucination in the action generation is a huge accomplishment for us, since it ensures that our game is variable while preventing false generations.

We were especially proud of our integration of generative AI and ground-up classification to achieve variable actions and impacts based on current events, surpassing capabilities of current simulations.

What we learned

We learned a lot of coding subtasks, from writing up websites to creating games. We also learned a lot about integrating moving parts, like APIs and ground-up models into React, along with logistics for the simulation itself. Most notably, we learned about how to identify biases in data and see how they affect the outputs of the model. It is really difficult to procure a clean and comprehensive dataset, and even when you do it is very difficult to rid it of any intrinsic biases.

What's next for PoliticsForYouth

We do plan on bringing our existing features to improved accuracy and better fluidity. After that, we plan on expanding the features in our game, potentially representing countries and making it multi-player (almost like a more-complex Risk game). We also plan on expanding to a set of games adapted from existing games (like HeadsUp) under a similar theme.

Built With

• openai
• python
• react.js