Anti-discrimination Insurance Pricing: Regulations, Fairness Criteria, and Models
With the rapid development of AI technologies and insurers’ extensive use of Big Data, a growing concern is that insurance companies can use proxies or develop more complex and opaque algorithms to ‘legally’ discriminate against policyholders. A legal grey area has resulted from this phenomenon – direct discrimination is prohibited by laws, but indirect discrimination can be tolerated without restrictions. This paper aims to establish the linkage of various insurance regulations, fairness criteria and anti-discrimination pricing models. To this end, this paper reviews anti-discrimination laws and regulations of different jurisdictions with a special focus on indirect discrimination in general insurance. It summarises different discrimination definitions and fairness criteria originated from both insurance and machine learning fields. Empirical analysis using a general insurance dataset is conducted to compare different anti-discrimination models and their impact on insurance pricing.