This article by Lemonade co-founder and CEO Daniel Schreiber tackles a profound issue for insurance and offers an innovative solution. The article suggests a smart way to watch for bias hidden in algorithms and to correct for it. In the process, Daniel provides an opening toward a holy grail: being able to price risk accurately for each individual.
The article is well worth your time. We're delighted to be able to share it with you and hope you'll share it, too. The change will require support not just from incumbents and insurtechs but also from regulators, whose structures, as Daniel notes, are reasonably friendly in Europe but would require more adaptation in the U.S.
I won't describe in any detail what Daniel calls his "uniform loss ratio" test, which makes sure that AI-based pricing for individuals produces defensible results for every group when losses are measured against pricing at the group level. But I want to build on his proposed test and explore the implications for how we'll all need to adapt to a world of much more individualized pricing of risk.
First, consider the technical requirements that must be met. Specifically, the data requirements will necessitate a continuous re-examination of privacy issues. The industry is already facing legislation designed to prevent an insurer's access to specific, individual data. A few in the public policy sector have taken this to an extreme by introducing legislation that would deny consumers even the option to voluntarily share data with their insurer for their own benefit.
Second, the more data that is aggregated by any organization, the more it becomes a target for bad actors. While all insurers ferociously protect their customers' data, the convergence of the required new computational capabilities and vast array of data raises the bar on cyber security significantly.
Third, basing premiums on an individual's risk profile will intensify the spotlight on operational expenses. As insurers zero in on an individual's risk, that individual will have more transparency about the process and will tend to sign on with whatever insurer can cover his or her risk at lowest cost.
Fourth, how will customers react? The move to individualized pricing creates huge opportunities for innovation, but consumers need to participate in the development. Would we not want consumers to have a choice between traditional, segmented, pricing and the new, individual pricing?
The benefits of individualized pricing are clear. If we can be sure to avoid bias, we can take advantage of the full array of capabilities of artificial intelligence. And the "uniform loss ratio" test can get rid of the "ghosts in the machine": biases that are unintentional but that are currently unrecognized and unavoidable given the limitations of our data and computational capabilities. We can then democratize access to services and products and accelerate the move away from ratings and recovery and toward preventing risks.
The journey from here to there:
- Will require a substantial collection of innovations,
- Will increase the clarity and urgency of certain issues.
- Rightly will drive a stake through the heart of discrimination,
- Represents an abundancy of opportunity,
- Is, let’s face it, inevitable.
Might as well get moving, right?
Chief Innovation Officer