AI and ethics: What advisors must know

Artificial intelligence is making inroads into many aspects of the insurance industry — from underwriting to onboarding to filing claims. But one of the dangers of AI is that it is expanding faster than the industry and individual advisors can keep up.

The financial services community must collaborate to address the risks of AI and promote ethical decision-making, said Azish Filabi, managing director of The American College Cary M. Maguire Center for Ethics in Financial Services. She has spoken and written on the ethical principles guiding the use of AI in financial services, emphasizing AI’s impact on clients and advisors.

Filabi told InsuranceNewsNet that confidentiality is a key concern when advisors use AI. 

“I think is important not to put any client information into a technology tool unless you’re absolutely certain what’s going to happen with that information, where the data is going and that the data is protected. So that’s one area of confidentiality that I think is important, not only from an ethical perspective, but it could have legal repercussions as well,” she said.

Confidence is another concern that Filabi said advisors must consider when using AI.

“Duty of care implies that you are knowledgeable about what you advise clients about and you take care that the information you give is accurate,” she said. She said advisors must be confident that they are inputting the correct information into AI tools and confident that AI is giving accurate answers.

“Generative AI tools are designed to be stochastic, which means that each time an input is put into AI, you might not get the exact same answer every time,” she said. “It’s not like using a calculator, where if you put in the same numbers, the calculator is designed to give you the exact same answer every time. Gen AI tools are more creative, and that’s the beauty of them. They can present different outputs that maybe you wouldn’t think of as a user. There are right and wrong answers in what advisors do. If the information isn’t accurate and you’re relying on it, that’s problematic both from an ethics perspective and a legal perspective.”

Filabi cited a court case from New York state in which an advisor was hired as an expert witness to advise on an evaluation matter relating to a mutual fund. 

“They used a generative AI tool to assist with the calculations — they were looking at valuation over a certain period of time — and the court used the same tool to kind of double-check the calculation, and they got different answers each time. And as a result, the court didn’t accept the evidence in that case,” she said.

Filabi said she believes this case is an example of how an advisor’s use of AI as part of their process “is introducing a new risk that could have repercussions down that line that maybe you’re not thinking about today.”

“This not only poses reputation risks for the advisor, but it has an impact on the client as well.”

The Center for Ethics in Financial Services has been studying AI since 2020, a few years before ChatGPT was released to the public, marking the beginning of AI tools being widely available to the public.

“We started studying AI from the perspective of the ethical risk that AI could present in underwriting,” Filabi said. “We began studying AI at a time when there was a lot of emphasis on diversity and inclusion issues, and we wanted to make sure that processes are fair, unbiased and inclusive.”

From there, researchers began to study how to support ethical decision-making when AI is used in the industry.

“We focused on how industry leaders and regulators can think about AI as it’s embedded into things such as underwriting technologies,” Filabi said. “We provided a three-pronged framework that included audits, testing and certification that contemplated the different players in a chain of AI accountability and the role that each of those prongs might have in making sure that the data is accurate and there are no mistakes and that certain ethical observations have been embedded in the process.”

Understanding AI’s limitations

It’s important for advisors to understand the limitations of AI in order to use it ethically, said Eric Ludwig, director of The American College Center for Retirement Income.

“You need to understand the knowledge component of AI,” he said. “These large language models are not necessarily databases of facts. They’re not like a big encyclopedia that spits out facts back to you. Their job is to generate text, and they are really trying to predict what the next word is in a sequence based on other patterns and probabilities from the data they were trained on. After someone understands that, then they can start to understand what those limitations are, and it goes back to the individual to verify that the output generated by AI is accurate.”

Advisors who use AI also must understand where the information used by AI goes.

“If an advisor is using these tools, they must realize that if they’re inputting personally identifiable information into it, that information could now become part of the large language model, which also means that it could potentially be extracted by someone else,” Ludwig said. “So it’s important to understand where the knowledge or data is going.”

When advisors use AI to determine the type of advice they give a client, are they obligated to inform the client that AI was used as part of the process? “My gut feeling is that they are not obligated to do so,” Ludwig said. He explained further.

“Let’s say I’m sending an email to someone. I don’t disclose that I used spell-check, which is a type of AI. Just the same as if I ran a Monte Carlo simulation using Microsoft Excel or some other program.

“So I think it still goes back to the advisor having to be the one who is responsible for verifying whether the output is accurate. They are using the AI tool to generate some time efficiency, which is totally normal. That’s why we use technology in the first place.”

Susan Rupe is managing editor for InsuranceNewsNet. She formerly served as communications director for an insurance agents' association and was an award-winning newspaper reporter and editor. Contact her at [email protected].