Deep Dive on Racial Inference Methods for Insurers

Insurers are increasingly expected to test that their use of external consumer data or AI does not lead to unfair discrimination of protected race or ethnic classes. But, race or ethnic class for insureds is not commonly known. Protected class membership must be inferred.

To date, focus has been on testing methodologies and not on inference of race or ethnic class inputs. However, the choices involved in racial group inference can have major downstream impact.

We seek to deepen the conversation around racial inference for insurers by:

• Giving background on the most commonly used methods

• Identifying key decisions involved

• Detailing best practices and offering intuition where the optimal course of action is less clear

Find the full analysis here and summary here.