Defining Fairness

What constitutes a "fair" algorithm in the criminal justice system?

In order to develop more fair algorithms, we began by defining what "fair" means in the context of risk assessment algorithms. Our qualitative definition of fairness is shown above.

We quantitatively define fairness as minimizing disparate impact among marginazied groups. For more information on how we used disparate impact to measure the fairness of each of our algorithms see our fairness repository on GitHub.

See our code on GitHub