Artificial intelligence is a valuable tool for insurance companies, but transparency and morality are necessary.
Key takeaways
- AI usage is becoming increasingly common in the insurance industry
- Ethical concerns are arising because of greater use and the methods employed
- Many algorithms use inconceivable amounts of data, making decisions impossible to explain
- Focusing on explainable models could effectively address AI ethics issues
Although the insurance industry’s reliance on data is nothing new, AI offers new ways to use it. Gathering customer information is incredibly quick and easy with artificial intelligence, helping insurance companies improve efficiency, reduce expenses, and optimize customer service.
These improvements make AI seem like a surefire benefit for insurers and customers, but there are ethical concerns insurance agencies must address. For example, using AI to predict customer behavior could create unfair bias as customers search for policies. It can also make it impossible for insurers to clarify offers presented to customers.
Finding a balance between the benefits of using this technology and AI ethics will be vital moving forward as stakeholders in this industry collect more customer data. Policyholders could also demand accountability as insurers use more of their information.
Transparency and ease of explanation will be essential for insurance companies using artificial intelligence. Customers will demand to know what information insurers are collecting and how they’re using it. Here are insights insurance companies can use when addressing AI’s ethical concerns.
Why transparent models are crucial
All insurance companies need transparent AI models. This means customers know and understand the information the model is collecting, and how it will use that information to reach policy decisions.
A lack of transparency and explanation present significant problems as the insurance industry adopts AI. For instance, many algorithms use black-box models to form outputs that customers can’t understand.
A black-box model could use thousands, millions, or even billions of parameters to create outputs. Human beings can’t comprehend such models; cognitive load theory suggests people can only understand models with between five and nine rules. So, it’s impossible for customers to know how the insurer create policies in this scenario, creating an ethical problem.
In comparison, a white-box model only uses a few rules to create a particular customer’s risk profile, ensuring the customer understands the policy’s pricing.
The argument in favor of using a black-box model is greater accuracy from additional data points. However, a study published in the Journal of Big Data suggests black-box models are more accurate in 69% of cases, but the differences are minor most of the time.
Further, the Journal of Big Data study reports that adding another step with surrogate modeling improves white-box accuracy, further reducing the difference. It concluded black-box modeling collects a significant amount of unnecessary information, producing a lot of noise.
In addition, using black-box modeling algorithms could hinder customer relations. If customers don’t understand how you’ve made policy decisions, they could be less likely to return. This could be increasingly likely as AI ethics become an insurance industry focal point in the coming years.
How to explain AI decisions to customers
The next step when addressing AI ethical concerns is providing customers explanations for policy outcomes. If employees can’t tell your customers the reasons behind their premium amounts or underwriting decisions, the company could face problems retaining these customers.
Explainability means your company provides an explanation you can present to customers so they can understand. Many AI systems don’t break information down into digestible formats, leaving the customer in the dark.
For instance, imagine a scenario where a customer asks for a car insurance quote through an online, AI-powered tool. That quote ends up far more costly than the customer’s previous policy, and the customer demands an explanation on how the system came up with that number.
If you don’t prioritize clear explanations when developing an AI model, you can’t tell the applicant how the system assessed their risk profile. You could end up facing discrimination and bias accusations, so doing everything possible to provide explanations for policy outcomes is vital.
Understanding the processes your artificial intelligence system goes uses makes it possible to better explain the system to potential customers. As people become more aware of AI’s prevalence AI in the industry, they will want more information on how each firm uses it. Developing a system that’s easy to explain to your clients reduces the chances of them avoiding your company because it uses AI.
Be forthcoming with customers about your AI usage, the data it collects, and how you’ll use that information when creating policies. That way, you can avoid many AI ethics concerns the industry is sure to encounter.
Balancing accuracy and transparency
As your insurance company increases its reliance on AI, find a balance between accuracy and transparency. You don’t want to pass up the benefits AI offers service providers, but you also don’t want to scare customers off or, worse yet, face accusations of bias or discrimination due to a flawed model. Fortunately, it’s possible to create AI models that factor in explainability, which allows transparency while taking advantage of the time savings, cost reductions, and accuracy improvements AI provides.
NICRIS Insurance Agency offers personal, life, and commercial policies in New York State. We understand how AI is transforming the insurance industry and spend time addressing ethical concerns surrounding AI whenever possible. Contact NICRIS Insurance today for more information or to get a quote.