From the course: Introduction to Auditing AI Systems
Join today to access over 23,800 courses taught by industry experts.
Statistical parity
From the course: Introduction to Auditing AI Systems
Statistical parity
- [Instructor] There are various metrics we can use to help us find disparities in AI system outcomes. These metrics evaluate different aspects of fairness in predictive models, but in some cases, they can contradict each other due to inherent trade-offs and assumptions between them. You should know that there are at least 21 quantifiable definitions of fairness and careful consideration should be taken to decide which definitions are best aligned with organizational goals. Now, let's cover some of the metrics we use for measuring fairness. First, we can use the equal opportunity metric to assess whether a model provides an equal chance for all individuals regardless of their membership in a particular group to receive a positive outcome. Let's say we have a model that predicts whether a job candidate will be successful in a role. In this case, the positive outcome means the candidate gets the job. Our goal is to have a…
Practice while you learn with exercise files
Download the files the instructor uses to teach the course. Follow along and learn by watching, listening and practicing.
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.