Like in many other sectors, personal data and algorithms are increasingly being used by insurers in their underwriting, claims, and rating practices. Whilst these advances can bring benefits, concerns have been raised about the quality of information sources and the rationale for using algorithms, particularly because insurance practices are high-risk and minority groups could be especially vulnerable to discrimination.
In response to these concerns, Colorado’s General Assembly enacted legislation last year that restricts insurers’ use of ‘external consumer data’, prohibits data, algorithms, or predictive models from unfairly discriminating, and requires insurers to test their systems and demonstrate that they are not biased. Here are 10 things you need to know about this legislation.
Insurers are prohibited from unfair discrimination in insurance practices and the use of external customer data and information sources or algorithms and predictive models that unfairly discriminate.
Insurers are required to i) outline the type of external customer data and information sources used by their algorithms and predictive models; ii) provide an explanation of how the external consumer data and information sources, and algorithms and predictive models are used; iii) establish and maintain a risk management framework designed to determine whether the data or models unfairly discriminate; iv) provide an assessment of the results of the risk management framework and ongoing monitoring; and v) provide an attestation by one or more officers that the risk management framework has been implemented.
A computational or machine learning process used to inform human decision-making in insurance practices.
A process of using mathematical and computational methods that examine current or historical datasets for underlying patterns and calculate the probability of an outcome.
A data or information source that complements an insurance practice or provides lifestyle indicators. Examples of sources include credit scores, social media habits, purchasing habits, home ownership, educational attainment, locations, occupations, licensures, civil judgements, and court records.
Unfair discrimination occurs when external customer data and information sources or algorithms or predictive models correlate with protected characteristics (race, colour, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity or gender expression) and result in a disproportionately negative outcome for these groups that exceeds the reasonable correlation to the underlying insurance practice (in respect to losses and underwriting costs etc.).
Insurers are not required to collect data relating to race, color, national or ethnic origin, religion, sex, sexual orientation, disability, gender identity, or gender expression.
Following a stakeholder engagement process, the Colorado Insurance Commissioner will establish bespoke rules for specific types of insurance and insurance providers, in relation to assessing and providing evidence of whether unfair discrimination is occurring. These rules will also likely establish penalties for non-compliance and a reasonable period of time for insurers to remedy unfair discrimination.
The legislation does not apply to title insurance, bonds executed by qualified surety companies, or insurers issuing commercial insurance policies. It does apply to insurers that issue business owners’ policies or commercial general liability policies if these policies have annual premiums of $10,000 or less.
The legislation will come into effect on 1st January 2023 at the earliest.
Holistic AI can help your firm:
Speak with us to find out more about how we can support your journey towards compliance.
Despite, AI specific regulation on a federal level being nascent, firms in the US should heed with caution as a precedent for civil liability is slowly being built. Recent examples include an ongoing lawsuit against insurance company State Farm; where it is alleged that their automated claims processing has resulted in algorithmic bias against black homeowners.
More recently, the State of Michigan reached a $20 million USD settlement with Michigan residents wrongly accused of fraud by an automated system used by the state. State governments and agencies are not letting this go unnoticed. For example. in New York, the Department of Financial Services reserves the right to audit and examine an insurer’s underwriting criteria, programs, algorithms, and models, to ensure algorithms are not in breach of existing law and insurance regulations.
However, proposals from the federal level are also being seen. For example, the Algorithmic Accountability Act is a proposed federal law that would require companies to assess the impact of the automated systems they use and sell in terms of bias and effectiveness.
Shaping up to be the body with the most hunger to regulate AI in the US, Bloomberg has predicted that 2023 will be marked by a determined, and aggressive FTC. Backed by its three-for-three records in getting settlement orders against companies that were investigated for their use or development of algorithms through suspiciously acquired data, this prediction is likely to ring true.
Written by Airlie Hilliard, Senior Researcher at Holistic AI. Follow her on Linkedin.
Subscribe to our newsletter!
Join our mailing list to receive the latest news and updates.
Our automated AI Risk Management platform empowers your enterprise to confidently embrace AIGet Started