Understanding the Consequences of Machine Learning Models
Multiple governing bodies are starting to regulate the use of machine learning models. This means producers or users of the methods must regularly simulate how their models affect different demographic groups. Simulating the effects of a selection profile is something Hogan already does. Every selection profile produced by the Hogan Data Science team includes a simulation of pass rates for different racial and ethnic groups, genders, and age groups.
Automating Pass Rate Simulations
We want to produce these simulations annually, but they take time. We are working on automating the simulations to make them quicker and more efficient. This project will also allow us to quickly run other simulations, such as checking for potential shifts in clients’ pass rates over time or computing impact analyses when we update a norm.