A Flurry of Algorithmic Activity at the NAIC 2022 Fall National Meeting
Surveys
As the most prolific group, the Big Data and Artificial Intelligence (H) Working Group (Big Data WG) reported on its Artificial Intelligence and Machine Learning (AI/ML) state surveys for:
- Private passenger auto (PPA) insurance – The Big Data WG trumpeted the results from the 193 responses. The PPA survey addressed:
- How AI/ML is used by PPA insurers;
- Data used by PPA insurers in AI/ML;
- Whether insurers are providing additional information about data elements to consumers other than what is required by law. The Big Data WG noted that, although the number of reporting companies is lower than expected, the answers reported are almost unanimously “no” for each of the insurers’ respective operations, except for rating, which had about 32% of the responses reporting “yes”;
- Consumers’ ability to correct data. As to whether consumers have more opportunity to challenge or correct their specific data than is mandated by the federal Fair Credit Reporting Act (FCRA), many did not answer. Of those who did, about 50% said “yes” for rating and underwriting, 40% said “yes” for claims and marketing, 15% said “yes” for fraud detection, and less than 10% said “yes” for loss; and
- Governance practices as they relate to the NAIC’s Principles on Artificial Intelligence. The Big Data WG noted that “a sizable number” did not answer the questions.
The survey also asked about PPA insurers’ documentation practices.
- Home insurance – The Big Data WG heralded that this survey is in process; and
- Life insurance – The Big Data WG proclaimed that it was modifying the survey based on comments received on an exposure draft. The American Council of Life Insurers commented that the draft survey would be a polar vortex for insurers, as the questions were more numerous than earlier surveys. Consumer representative Birny Birnbaum believed the broad blanket of questions failed to obtain specific enough information about biases, testing for biases, and the use of biometics. He also asserted that the survey was foggy as it would not allow regulators to see whether the data and algorithms used by insurers are reasonable or unreasonable.
The Big Data WG also decreed that that 192 call letters will be sent out, of which six will go to insurtech companies. The Big Data WG plans to open its survey website and distribute initial call letters in January, with a formal call letter in February. After the formal call letter is sent, the companies will have 30 days to bundle up their responses.
The Big Data WG explained that the results from the three surveys would be used to evaluate whether changes should be made to regulatory frameworks.
Guidance/Framework
In addition, during the Big Data WG meeting, the Accelerated Underwriting (A) Working Group (AUWG) joined the fireside chat by stating that in the New Year, it plans to expose draft regulatory guidance for accelerated underwriting practices. It plans on collaborating with multiple working groups, committees, and the collaboration forum on the guidance document. Again, Mr. Birnbaum raised concerns about credit and biometric information being used in accelerated underwriting. He proposed that insurers should have to show there is no bias if they want to use biometric information in accelerated underwriting.
The Innovation, Cybersecurity, and Technology (H) Committee announced that, to freeze the ability for AI to result in illegal bias and discrimination, it would be developing a regulatory framework on algorithmic bias. This regulatory framework will be in the form of a model bulletin, which will include the following sections: background, definitions, regulatory expectations, and regulatory oversight and examination standards. A number of working groups that are part of the “collaboration forum” will assist with drafting.
***
As the NAIC continues to explore consumer data and AI/ML, forecasts show that these Fall Meeting flurries could develop into a blizzard as the New Year develops.