Ethical Use of AI & External Data in Insurance

Insurers are increasingly turning to artificial intelligence (AI)-enabled analytical models and external consumer data sources to enhance risk management, enrich decision-making, increase operational efficiencies and improve financial performance. At the same time, regulators worldwide are intensifying their focus on the governance and fairness challenges presented by these complex, highly innovative tools, particularly the potential for unintended bias against legislatively protected classes of people. Companies need a clear plan and uniquely skilled and experienced resources to ensure that they responsibly use this rapidly advancing technology.

What is Ethical AI?

AI has the potential to amplify and scale human biases at an unprecedented rate, sometimes in an unintentionally harmful way. AI tends to obfuscate how it arrives at a given decision, therefore it can be a challenge to manage performance and understand its “decision-making.” Explain ability testing and documenting seeks to eliminate the ambiguity that exists around AI model development and training, data consumption, and algorithmic outputs.

To establish clear lines of responsibility for the ethical use of AI and third-party data, and ensure compliance with evolving regulations, insurers should leverage both existing internal governance structures, including regulatory and legal teams, and experienced independent professionals.

Expanding the scope of internal governance and legal teams to include oversight of ethical AI priorities is a natural extension of their current remit. However, supplementing this with the engagement with independent experts to test and validate model efficacy will strengthen interactions with regulatory bodies.

A Sample of Global Regulatory Entities Currently Reviewing the Use of AI*

U.S. State Departments of Insurance (e.g., Colorado, California, New York, Connecticut, and New Jersey)
National Association of Insurance Commissioners (NAIC)U.S. Federal Trade Commission (FTC)
U.S. Department of Commerce (National Institute of Standards and Technology or NIST)
U.S. Consumer Financial Protection Bureau (CFPB)
U.S. Department of Justice (Civil Rights Division)
U.S. Equal Employment Opportunity Commission (EEOC)
European Union (EU)
United Kingdom Financial Conduct Authority (FCA)

* As of mid-2023

AI Governance and Bias Assessments

To identify and remediate unintended model biases or defend outcomes that may be reasonably unavoidable, independent model testing by qualified professionals can be embedded as a core component of an insurer’s plan for corporate responsibility, regulatory compliance and effective risk management.

FTI Consulting’s interdisciplinary approach helps clients proactively and reactively assess the transparency of AI solutions and identify and remediate potential biases. We apply advanced statistical methods, proven end-to-end proprietary frameworks and accelerators, data science and technology know how, and deep industry expertise and assist clients to:

Understand how AI is being used across the organization in such functions as marketing distribution, underwriting, pricing, claims and fraud management;
Identify any biases or unintended consequences that the use of AI maybe causing;
Develop strategic and data-driven remediation plans; and
Identify opportunities for improvement that will result in operational, customer retention and growth, and financial performance gains over time.

Our team of seasoned data science, technology, regulatory, actuarial and insurance operations experts perform all model and data analyses and assessments in a fully controlled environment that leverages best-in-class security protocols. The scope of an assessment varies based on the individual needs, but can include any or all of the following:

AI Governance & Risk Management:

Assess AI and “big data” processes, policies and procedures
Data and information management
Model validation practices

Data Assessment

Analysis of underlying data sources
Feature engineering, selection and importance

Qualitative Model Assessment

Model-specific governance, policies and controls
Model lifecycle management (e.g., model selection, training, validation, approval and monitoring)

Quantitative Model Assessment

Disparate Impact: Using benchmark data determine if the distribution of protected classes in the output of a given model is materially different from that of the benchmark data (e.g., census) to determine if the model is disparately impacting certain types of individuals
Proxy Variable Identification: test whether a given attribute provides the independent predictive capability for the use case that is separate from protected class associations

Explainability Assessment and Support

Evaluate and document existing data inputs and outputs for all AI-enabled models across all business functions
Propose enhanced explainability strategies in advance of potential audits, inquiries and/or legal challenges
Identify, assess and explain circumstances of commercially unavoidable bias

Final Report and External Communication Support

Formal findings and remediation recommendations as appropriate
Preparation of internal team for regulatory review and/or other external inquiries or challenges

Why Choose FTI Consulting?

FTI Consulting’s Global Insurance Services practice is driven by more than 100 insurance professionals with deep industry experience and knowledge, business acumen, and technical and leadership skills. We combine this industry expertise with complementary skills from other segments of our business to create fit-for-purpose, multidisciplinary teams to address client challenges and opportunities, including the need for AI model bias governance, assessment and remediation support in the face of new and expanding regulations.

Our Data & Analytics team is constantly evolving with the needs of our clients – leveraging cutting-edge tools and techniques to reveal and enrich analytical insights. We are technology agnostic and leverage our clients’ preferred tools and platforms to drive insight and impact. Building on our deep industry experience and technical expertise, we empower clients to optimize outcomes in critical events and evolving matters.

Let us show you how we can enhance your organization’s response to rapidly evolving industry guard rails and regulatory mandates for your use of AI and third-party data. We understand the challenge and are uniquely positioned to help you address it – both now and moving forward.