Benevolent Bots is Lemonade’s new podcast series about the overlap of a few issues that are always on our minds: Artificial intelligence, insurance, and ethics.
Co-hosted by Lemonade CEO Daniel Schreiber and our in-house AI Ethics and Fairness Advisor Tulsee Doshi, Benevolent Bots takes a deep dive into big questions about technology and responsibility. Our third episode features a conversation with Scott Fischer, partner at DLA Piper.
Prior to his private sector work he served on the regulatory side of things, as Executive Deputy Superintendent for Insurance at the Department of Financial Services. He’s also the former Deputy Superintendent at the New York Liquidation Bureau, which is “where insurance companies go to die,” he tells us. Yikes.
“Today we’re discussing the complexities of AI regulation. What do regulators need to be thinking about? Why is this so challenging? And how do we actually determine standards for questions like, ‘What is fair’?”Tulsee Doshi, Lemonade’s AI Ethics and Fairness Advisor
Now, AI ethics is a nuanced field. It contains multitudes! And it’s very much not suited to quick soundbites. Topics covered in this episode include credit scores and insurance proxies; whether lawyers really get consumers; thorny regulatory problems as both tight rope and mine field (watch out!); and why a little bit of uncertainty can be productive.
If you’re looking to quickly glean some insights before fully committing to the episode, we’ve summarized a few key points below (edited and condensed for length and clarity).
Let’s clear one thing up first
“Before you learn anything about insurance, you think, ‘Oh my God, it couldn’t possibly be more boring,” Scott Fisher says. “And then you get into it. And you’re like, ‘Wow, this is really interesting. I mean, it’s not sexy, but it’s very interesting.” We’d like to clarify here that Lemonade’s official position is that insurance is both interesting and sexy.
Regulators are from Mars, VC investors are from Venus…
“There is a significant disconnect between [regulators] and folks on the insurance side, particularly in what is called insurtech,” Fischer says. “I don’t mean this in a critical or negative way whatsoever. It’s just a very different mindset. What I’ve been able to do in private practice is try to bridge that big gap between what the regulators expect, and what business wants. Quite frankly, a lot of times within the insurtech space you’re talking about people who are backed by VC or private equity—people that are really not in that mindset.”
That doesn’t mean the relationship has to be fraught. As Fischer implies, communication and understanding can go a long way. “Insurance regulation in the U.S. is different than other kinds of regulations,” he says. “It’s very hands on, it’s very touchy-feely. Although it is rules-based, I think everybody in the industry would say—and I don’t think this is controversial—’The rules are not as hard and fast sometimes as you might think that they are.'”
Slow and steady?
“Insurance has been based on science [that was] high-tech at the time, and is low-tech now. Actuarial science. Regulators, are very much accustomed to dealing with that, that’s how they feel comfortable, that’s what they’re doing,” Fischer says.
“AI is sort of changing all that. There’s tension that gets created. [Regulators] don’t know exactly how it works. And the framework in which they are supposed to operate—the laws that are written—are not designed with that in mind.”
“It may feel to those in the insurance community—or insurtech, or anybody that’s coming to insurance now—that there’s this rigidity or slowness. It’s there, but it’s there because it’s a function of the way that the system has been built over time. And it’s going to take a while. It’s going to take a lot of effort to move beyond just basing everything on the current state of actuarial science.”
Embrace a little imperfection
“As consumers, as business people, as participants in the market, we have a right to expect our regulators and legislatures to grapple [with these issues]…. to create reasonable expectations,” Fischer says.
“There is going to be imperfection here. And [don’t] be afraid of that imperfection, because frankly —it’s already there…. The actuarial science that we have today—as good as it is, it clearly embeds biases. It clearly embeds inequities.”
“[Regulators] shouldn’t be afraid of—I don’t want to say, ‘breaking a few eggs’—but coming out and saying: ‘This is what unfairness looks like.’ And therefore giving the AI experts the opportunity to say, ‘Okay, great. We understand what you’re talking about now. Let us try to work within that construct. Because right now we’re going to do what we’re going to do, and we have no guardrails.'”
The awful and great United States of insurance
“One of the awful things and the great things about insurance regulation in the U.S. is it’s so diffused. There’s 50 states, D.C., and five territories,” Fischer explains. “They can be right next to each other, New York and New Jersey, or New York and Connecticut, New York and Pennsylvania. Different rules. What happens if you allow for AI to be used in one respect, one state, but not in the other? What happens if it actually turns out that auto insurance becomes more expensive in the state that allows for this unfettered use of AI? I mean, what then?”