Recent regulatory developments in states like Connecticut (CT), Vermont (VT), New Hampshire (NH), and Alaska (AK) have seen the adoption of the NAIC AI model bulletin. The bulletin underscores a growing emphasis on ensuring compliance with anti-discrimination laws and promoting fair underwriting practices and fair practices generally along the consumer journey when the carrier is using external data sources and different types of complex models.
While the model bulletin provides a framework for regulatory oversight, it's noteworthy that they are still working towards the specifics of testing methodologies. However, states like New York (NY) and Colorado (CO) have taken proactive steps to address this aspect.
In New York, a DRAFT Circular Letter has been issued, emphasizing the importance of testing and analysis to detect any potential biases or disparities in underwriting decisions. Insurers are encouraged to use multiple statistical metrics, including adverse impact ratio analysis and standardized mean differences assessment, to evaluate the fairness of their underwriting practices.
Similarly, Colorado has proposed regulations that outline stringent testing requirements for insurers. These regulations mandate insurers to conduct quantitative testing of their underwriting processes, utilizing statistical methodologies such as logistic regression and linear regression. Insurers are required to assess whether underwriting decisions exhibit disparities based on factors such as race or ethnicity and take remedial actions as necessary.
CO looks and feels a lot like disparate impact testing while it does not expressly say it while NY is pretty clear as to that matter.
As more states consider regulatory measures in this space, it will be interesting to observe how they address the testing component and whether there is convergence or divergence in approaches across jurisdictions. Ultimately, the goal is to foster trust and confidence among consumers while the carriers meet their regulatory obligations.
Comments