JD, CPA, AEP®
Subscribe to Newsletter
April 01, 2022
Interest among state-level lawmakers in the potential for unfair discrimination due to artificial-intelligence-driven algorithms and predictive modeling techniques is on the rise.
On March 5, 2022, we had the privilege of presenting at the National Council of Insurance Legislators (NCOIL) Spring Meeting relating to the challenges of regulating artificial intelligence (AI)-enabled underwriting for life insurance. There’s been an increase in the use of AI in various insurance processes, including marketing, risk classification, and underwriting. As we’ve written before, these developments can create new opportunities for the industry, including increased access and improving pricing for consumers. However, they also create risks—particularly with respect to discrimination.
NCOIL is an organization comprised of legislators serving on state insurance and financial institutions committees around the U.S. In addition to serving as an educational forum for policymakers and the public on these topics, they write model laws on emerging topics from which States can benefit. We presented to the Financial Services and Multi-Lines Issues Committee, presided over by Vice-Chair Jim Dunnigan—a member of the Utah House of Representatives.
Our presentation was based on a paper we recently published in the NAIC Journal of Insurance Regulation, AI-Enabled Underwriting Brings New Challenges for Life Insurance: Policy and Regulatory Considerations. Based on our research, we proposed in the paper a multi-pronged governance approach to address the novel data sources, systems, and related risks of AI-driven insurance systems.
The first prong is to establish industry standards (derived from actuarial best practices) that are calibrated to the algorithm risks associated with the process. For instance, these standards could address accuracy in the data, the level of actuarial significance expected for data input, and related outcomes measures for how an algorithm performs. Prongs two and three of the governance framework require testing of the algorithm vis-à-vis the standards created. We believe the algorithm should be tested thoroughly, and that it should be certified to adhere to the agreed-upon standards before it is deployed. After it has been in use for a period of time, it should again be tested to determine whether it performed as expected.
A number of States have already addressed AI-enabled underwriting in insurance. Colorado passed an act in 2021 mandating that the state’s Insurance Commissioner establish rules requiring insurers to demonstrate technology, such as algorithms and predictive models, don’t unfairly discriminate based on race, color, national or ethnic origin, or a number of other protected categories. In Rhode Island, the state’s House of Representatives is considering a similar law that prohibits unfair discrimination in insurance processes when using external data sources or predictive modeling and algorithms.
The New York Department of Financial Services also addressed this topic in 2019 through a Circular Letter that expressly prohibited using criteria for underwriting purposes unless the insurer can establish the approach is not unfairly discriminatory pursuant to existing rules.
Our presentation was one of several sessions related to the impact of technology and related regulation on the insurance industry, demonstrating NCOIL’s commitment to lead regulatory efforts in this area. We are proud to have been a part of the discussion and look forward to continuing our work.