Colorado Division of Insurance Continues to Navigate Use of ECDIS and Algorithms by Life Insurers | Carlton Fields
On July 8, the Colorado Division of Insurance (CO DOI) piloted its third stakeholder meeting where the CO DOI continued to chart the underwriting practices of life insurers as part of the rulemaking process. required to implement SB-169 (as codified in Colorado Statutes Section 10-3-1104.9). The CO DOI discussed its proposed upcoming survey and data call for life insurers to gather information that will inform Colorado’s regulatory development. In addition, during the meeting, the ACLI presented its “ACLI Proposed Algorithmic Accountability Regulation”.
- Industry Survey and Data Call
At the July 8 stakeholder meeting, the CO DOI announced that it would release a survey and call for data from a sample of life insurance companies. The CO DOI proposed that the top 10 carriers with the most lives covered be invited to respond to the survey and call for data. The CO DOI has requested comments on its proposed survey and data call by July 22, 2022, and expects to release the survey and data call in mid-August. The CO DOI indicated that it would hold another stakeholder meeting before the survey and call for data is launched, but also stated that it may be possible to respond to comments without the need for a meeting.
The CO DOI believes that the survey and call for data will provide it with an impact on:
- Current practices in the use of ECDIS and algorithms.
- Views of the various components of the SB-169.
The proposed survey will gather information about a company’s governance practices and unfair discrimination testing process. The survey questions seek to determine what insurers are doing to assess whether their use of predictive algorithms or models using ECDIS results in a disproportionately negative outcome for protected classes that exceeds reasonable correlations with insurance practices. underlying. In particular, the survey aims to obtain information regarding four categories, as follows:
- Processes and Procedures Regarding Unfair Discriminationincluding (i) whether there are specific external data sources or types of external data that the insurer does not use, as part of the company’s internal policy, and (ii) which of the sources of external data, predictive models or insurance practices of the insurer are currently being tested for unfair discrimination.
- Evaluations of external data sources – i.e. inputs from predictive algorithms or modelsincluding how the insurer decides whether a given ECDIS should be tested, what bias is tested, what testing methodology is used and what analysis is used to assess a given ECDIS.
- Evaluation of all predictive modelsincluding what results are considered in tests of predictive models, what bias is tested, what testing methodology is used to determine race or other demographic information, what analysis is used to assess results, e.g. there is a disproportionately negative outcome for certain demographic groups, and what steps are taken to mitigate issues once they are discovered.
- Documentation of actual instances of governance and test procedureswhich seeks to obtain actual documentation of the governance and testing procedures an insurer has conducted and information as to whether or not there has been a finding of unfair discrimination.
In addition to the proposed survey, the CO DOI is proposing to launch a data call to determine if insurers’ data would be navigable for stress testing to determine if there is unfair discrimination in their use of ECDIS. The proposed data call seeks information on:
- Application data for products that use ECDIS in one or more predictive algorithms/models
The proposed data call seeks specific information regarding applications received during a given period with a minimum of 10,000 applications for life insurance products that use models incorporating ECDIS in the underwriting process.
- List of products that do NOT use ECDIS in at least one predictive algorithm/modelas well as a brief explanation of the data/information used and why it falls outside the scope of the law.
Birny Birnbaum suggested that the CO DOI consider reducing the veils by limiting the investigation and data call to term or whole life, as it is products sold in underserved markets that present the greatest risk. higher of unfair discrimination.
The CO DOI emphasized that it would not use information from the investigation or the data call to drag an insurer into enforcement action.
- ACLI Proposed Rules on Algorithmic Accountability
As the stakeholder meeting began, Jillian Froment, former Ohio Insurance Director, presented the ACLI’s Proposed Algorithmic Liability Rule (ACLI Proposal). ACLI’s proposal adopts a risk management framework that addresses unfair discrimination in underwriting based on race. The risk management framework would require insurers to:
a) develop and maintain an internal governance process regarding the use of ECDIS in underwriting to mitigate the risk of unfair discrimination;
(b) prepare a report documenting the insurer’s application of its internal governance process regarding the use of ECDIS in underwriting to mitigate the risk of unfair discrimination;
(c) perform and maintain impact assessments (IAs) to determine whether the use of ECDIS in a predictive algorithm or model in underwriting results in unfair discrimination based on race; and
(d) prepare an impact analysis report documenting the insurer’s analysis of its use of ECDIS in its predictive algorithms and models under a two-part structure.
Under the principles-based approach, each insurer would determine the most appropriate position and course for testing to determine whether the insurer’s use of ECDIS results in unfair discrimination based on race that exceeds a reasonable correlation with the underlying insurance practice. Consumer representative Birny Birnbaum criticized this approach, saying it would force the CO DOI to review several testing methods as they would vary by insurer. Colorado Commissioner Conway expressed concern that ACLI’s proposal strayed a bit from shipping lanes because he disagreed with insurers determining method or threshold appropriate for testing unfair discrimination. He proposed that the Colorado regulations set the stage for the required tests.
The two-part reporting structure of the ACLI proposal would encourage insurers to first document their internal governance process, testing, methodologies and results (Part 1), then summarize the prior documentation and provide an attestation that they have carried out the tests in accordance with the regulations. (Part 2). ACLI’s proposal provides that an insurer would file both Part 1 and Part 2 in the first few years, and if the insurer’s use of ECDIS has not changed course year to year, the insurer could simply file Part 2 each year and make Part 1 available to the CO DOI upon request. The Commissioner expressed concern that given the rate at which the industry innovates and evolves in its use of ECDIS, there will always be changes to Part 1.
The Commissioner’s comments on the ACLI proposal, the proposed investigation and the call for data clearly indicate that the CO DOI has a different tact in mind than the ACLI proposal. We will continue to monitor the wind direction to see how the belayers should hoist the sails.