Inspiration

With the current events going on in the world when we decided to create a project based around racial equality. Everyday on the news we would see a new case of racial injustice with police officers and the judicial system, and we wanted to create an AI project that could help the problem.

When the theme was released we thought of creating an AI project which could decide whether or not a person being charged deserves bail or not. The theme required the project to be able to help a group of people in an historical event, and our project would be able to help the Black Panther Activist Group. Our project would've helped this group as 85% of the white lynch mob was released with low bail whereas only about 10% of the Black Panthers were released on bail and those released were only released on ridiculously high bail, making it near impossible for them to post bail. The Black Panthers group is explained more thoroughly in our website and our demo.

The main issue with bail is that the judge can decide whatever they want, and the judge will make assumptions about people. Although the majority of judges are not actually racist, everyone has some bias to them, and this can sway their decisions, however it shouldn't. Our program is able to solve this.

Another issue is that the datasets given from the court is bias, so the only true way to fix the issue is by creating a model which does not take certain inputs, such as race, and create an unbiased model.

What it does

Unbiased Bail is a progressive web application which allows Judges to decide whether or not a defendant deserves bail or not, without bias. A Judge will put in a csv file of the defendant, which contains all their information such as past convictions, juvenile felony count, and much more and the model will output whether or not the defendant should get bail. The model is unbiased, as it does not take into account race, sex, age and more features which should not be considered when deciding bail.

The project contains an SVM(Support Vector Machine)(Sklearn) model, which is a linear model. This model is great for us as we need it for a binary classification problem(bail or no bail).

How We built it

To build the base website we used HTML5, CSS3, and JavaScript. We used Javascript in order to create animations for the website, and help get the CSV file into the AI model. For AI model we used Python and Sklearn. We spent many hours testing different algorithms, and eventually came to the conclusion that the svm model produced the best results. Another model we tried extensively was a logistic regression model, however this did not produce the results we wanted. To move the CSV file into the AI model we use a PHP command line, and Javascript.

Challenges We ran into

The biggest challenge we faced regarded our model. It took quite a while to figure out which model is the best, and we figured this out by testing different models and comparing the accuracy. We also ran into issues with getting the file into the model, however we were able to switch our method to PHP, which allowed us to transition the file easily.

Accomplishments we are proud of

We are proud that we are able to create a project which takes a step forward regarding the current issues in the world. We feel accomplished as we were able to research, program, and test a project which can be used in such an important place.

What is next for Unbiased Bail

Our website is ready for the world. It is very scalable as our project is full functional, and can be used by any federal court or judge. Our website is necessary for the United States, as racial injustice will not stop and will always be seen as an issue unless something real is done about it. In order to create a public demo we need a domain which supports PHP, however we currently do not have this, so we are unable to host the site publicly. This website is a step in that direction and will be seen as an improvement to society.

Built With

Share this project:

Updates