Case study
Discrimination-free insurance pricing
Ensuring that insurance pricing is free from discrimination against protected characteristics such as ethnicity.
Challenge
Solution
Insurance companies often use unawareness pricing to avoid direct discrimination by excluding protected characteristics from model features.
However, indirect discrimination arises from correlation between protected and unprotected characteristics, meaning that even unawareness pricing is not free of discrimination.
Impact
Creation of a causal model of the variables used in the current model for price determination.
Identification of protected and unprotected characteristics and cases of indirect discrimination that were not explainable by resolving factors.
Creation of a causal pricing model that is free of both direct and indirect discrimination with respect to the specified protected characteristics.
Framework to detect indirect discrimination within existing insurance pricing models.
New causal pricing model that is mathematically proven to be free from direct and indirect discrimination.