Insurers cannot escape ‘bias testing’ going forward, speakers say

SAN ANTONIO -- "Bias testing" is a phrase insurers are getting acquainted with as artificial intelligence, big data and algorithmic programs become more commonplace.

They don't really have a choice because regulators are certainly looking at bias testing. For example, New York regulators are looking at governance protocols for AI systems and so-called external consumer data and information sources and employ fairness testing of predictive models and variables before they are put into use.

Currently, insurers aren’t required to abide by testing obligations when using AI in underwriting and pricing.

A session kicking off the Life and Annuity Conference, sponsored by LIMRA, LOMA, SOA and ACLI, tackled the topic Monday afternoon. Titled, "Avoiding Disparate Impact Through Enhanced Underwriting Models," the session focused on how insurers can meet regulators' expectations on the bias issues.

Julia Romero is a second vice president with Munich Re. She talked about how insurers can conduct fairness testing on underwriting models to check for disproportionate negative impacts on protected groups.

"When your model makes a mistake, what is the negative outcome or the negative consequence that it may have?" she asked. "This is kind of the foundational question that we need to ask when we're doing bias testing because it's going to inform some of the decisions that we make in terms of actually measuring bias, as well as guide us in the process of remediation."

States are acting fast on bias testing

Various states are adopting rules to govern AI and data use and they are not exactly the same rules, explained Kirsten Pederson of the American Academy of Actuaries. Different regulators are using different definitions and uniformity of rules is not something insurers should count on anytime soon, the panel agreed.

Colorado is likely the insurance regulators insurers are most concerned about, Pedersen noted.

The Colorado Division of Insurance recently opened a public commentary period for its proposal to introduce regulation to govern the use of AI in health insurance, as it investigates whether existing legislation can be applied.

In November, Colorado enacted Regulation 10-1-1, which applies to life insurance carriers in the state. Around that time, legislators also began looking at whether the regulation could apply to private passenger auto insurance.

On June 1, the regulation requires all Colorado-licensed life insurers to submit a compliance progress report. Next up is the testing component, Pedersen said.

"Colorado sent out a testing draft last year, received lots of feedback, and is now working on updating it," she added. "They decided to concentrate on outcomes, and so, premium and then your underwriting approval."

NAIC model gaining traction

While Colorado and New York are going their own way, many other states are embracing the Model Bulletin on the Use of Artificial Intelligence Systems by Insurers, adopted by the National Association of Insurance Commissioners in December.

The NAIC Model Bulletin on the Use of Artificial Intelligence Systems by Insurers is in place in Alaska, Connecticut, New Hampshire, Illinois, Vermont, Nevada, and Rhode Island. Other states are expected to follow in the coming weeks and months.

While the bulletin doesn't come with any mandates, it states that insurers “are expected” to develop, implement and maintain a written program for the responsible use of AI systems.

The model bulletin also encourages insurers to use verification and testing methods “to identify errors and bias” and the potential for unfair discrimination in predictive models and other AI systems.

The NAIC model bulletin includes a section addressing use of third-party vendors, Pedersen noted, which is important so that insurers cannot pass on responsibility for AI failures.

InsuranceNewsNet Senior Editor John Hilton covered business and other beats in more than 20 years of daily journalism. John may be reached at [email protected]. Follow him on Twitter @INNJohnH.

© Entire contents copyright 2024 by InsuranceNewsNet.com Inc. All rights reserved. No part of this article may be reprinted without the expressed written consent from InsuranceNewsNet.com.