Algorithmic Discrimination
Unfairly different treatment of individuals or groups by an automated decision system, often through proxy variables or biased training data.
Algorithmic discrimination occurs when an automated system produces unfairly different outcomes for people based on protected characteristics such as race, gender, age, or disability. The discrimination may not be explicit; it often arises when a neutral-looking variable, such as zip code or credit history, correlates with a protected class.
Colorado’s SB 24-205, later narrowed by SB 26-189, originally centered on algorithmic discrimination in high-stakes decisions, including insurance. The revised law focuses on disclosure and recourse rather than imposing broad risk-management duties. New York’s Circular Letter No. 7 and the NAIC Model Bulletin both require insurers to test for and document these effects.
The practical response is to screen inputs for protected-class proxies, test model outputs for disparate impact, and keep records of less-discriminatory alternatives considered. See our analysis of Colorado SB 26-189 and guide to proxy testing.