What Is Disparate Impact In Machine Learning?
Disparate impact is a quantitative measure of the adverse treatment of protected classes that compares the pass rate – or positive outcome – of one group versus another.
For example, a company with a fraud model might want to look at fairness of different groups. If disparate impact falls outside of the 0.8-1.25 range, it may mean that the sensitive group – such as people residing in zip codes where most of the population lives under the poverty line – is experiencing potentially discriminatory treatment. Clicking a level deeper might reveal disparate impact is the most pronounced for certain categories of purchases, such as for bail bonds, legal services, colleges, and even drugstore purchases.