Insuretech: Using AI to Eliminate Bias
Is Differential Pricing Fair?
Insurance companies are in the business of assessing and pricing risk. The insurance business has been around for hundreds of years but for much of this time insurance companies really did not know very much about some of the risks they were insuring, let alone how to price them efficiently. Those of you who have read the book Against the Gods (Bernstein) will know some of this story, including the origins of longevity tables which revolutionized the pricing of annuities in the 18th century.
The insurance industry has come a long way since Edmond Halley (the comet guy) first published his “life table” in 1693, the same year that W&M was founded. But the advent of big data and AI promises even more revolutionary changes in the not too distant future. The caption of this post—”AI Can Vanquish Bias”—comes from a recent two-part article written by Daniel Schneider, the CEO of the insurance company Lemonade Inc. (Don’t ask me about the name of the company; I really have no idea where it comes from.) The article was sent to me by Thomas Smith, former head of sales and marketing at Markel, and a Beyond Banking subscriber, who is known to many of you as a dynamic and engaging speaker in my Financial Services course at W&M. Thank you Smitty!
You can read Mr. Schneider’s brief article (Part 1 of a two-part series) here: https://www.linkedin.com/pulse/ai-can-vanquish-bias-daniel-schreiber/
Mr. Schneider’s article raises a number of interesting and important questions about the business of insurance in the era of big data and AI. But it also raises some more fundamental questions that we as a society need to think about and address in the context of insurance company regulation. Specifically, we need to think about the social tradeoffs between “efficiency” and “equity” in a world of individually differentiated price discrimination by insurance companies, enabled by big data and AI. (I am using the word “discrimination” here in a positive sense, one that will be very familiar to those of you who are or were Econ majors).
It may well be true, as Mr. Schneider argues, that the enhanced price discrimination enabled by big data and AI will actually reduce insurance bias in some important ways, by pricing insurance policies in line with the policyholder’s individual risk characteristics rather than the average risk characteristics of the class or classes to which the policyholder just happens to belong (gender, race, age etc). And this outcome may well be good from the perspective of economic efficiency. But is it “equitable” or “fair”? And even if you personally think it is equitable and fair, will our society tolerate this sort of price discrimination and/or the use of individuals’ “private” data for these sorts of commercial purposes?
In Part 1 of his article, Mr. Schneider provides us with a hypothetical example of price (premium) differentials in homeowner’s policies which discriminate between religious and non-religious policyholders, based not on their religious beliefs or practices per se but rather on the policyholders’ propensity to burn candles (for religious purposes) and thereby risk setting their homes on fire. The “efficiency” side of my brain accepts this sort of price discrimination as rational and to be encouraged. And even the “equity” or “fairness” side of my brain is probably ok with this, although I love candles and would light my house with them if I could. (My kids say that I should have been born in the 16th century.) But I have serious doubts whether American society at this point in time is really prepared to accept this sort of (extreme?) differential pricing, which on the surface really does look as if it is targeting religious observers (and in Mr. Schneider’s example, Jews specifically.). Just think of the reaction on Twitter!
I’d like to end this post here, but I plan to follow up soon with another one addressing some of the issues raised in Part 2 of Mr. Schneider’s article, in particular the issues of “fairness” and “equity” alluded to above.
PS: I have said this before and I will say it again: Insurance is a really interesting business and it’s going to become even more interesting in the years to come. Those of you who are quantitatively inclined should check it out. Lots of good jobs available, with the prospect of rapid promotion as all the old-fogies now running these companies retire or die off.