Sears Merritt is one of the people helping life insurers figure out what applying fast computers, giant databases, and rocket-scientist-type math to insurance can do for performance.
Merritt has been the chief data scientist at Massachusetts Mutual Life Insurance Company since 2015, and he is one of the fathers of the artificial intelligence (AI) technology inside the silicon brains of Haven Life, MassMutual’s web-based life insurance sales startup. MassMutual now has about 80 people on its data science team, up from none in 2013.
(Related: Two Online Life Distributors Expand)
Merritt has a bachelor’s degree in electrical engineering from the University of Colorado at Boulder, a doctorate in computer science from the University of Colorado, and a dissertation that provides mathematical models for understanding competition. To create the dissertation, he used collections of data that, in some cases, took up more than 80 terabytes or storage space, or the equivalent of the disk space available on about 400 typical laptop computers.
He’s now applying similar statistical analysis and database-tending skills to exploring the secret life of insurance underwriting data.
Life insurance underwriters and actuaries have always been able to look at one or two variables, such as weight and occupation, affect mortality. Graphs that show how one or two variables change over time are common.
The actuaries have had a hard time analyzing the effects of many different, possibly related variables simultaneously. To simplify their work, they have often assumed that key variables would stay the same over time, or change in a steady way over time.
A MassMutual AI system can now take large batches of data and create, what for the AI system, amounts to a high-resolution, AI-friendly graph, with a different dimension for each variable, and with the each variable plotted in a way that reveals each variable’s kinks and curves.
A human cannot see or think in enough dimensions to look at a many-dimensional graph the way an AI can see it. A well-designed AI can find multi-dimensional bubbles and tangles in the graph that may show how the variables interact with one another, and where real claims experience appears to be much different from what the actuaries’ old, human-friendly models predicted.
MassMutual’s data scientists then have to use statistical testing techniques to decide whether the apparent relationships or gaps are real, or the result of data or AI analysis problems.
One result has been MassMutual’s new LifeScore360 tool. The tool can take 40 types of traditional life insurance underwriting data from a user, convert the data points into a risk score, and spit out a graph explaining the risk score.
The list of companies using the LifeScore360 tool includes iPipeline, a life insurance technology company, and Swiss Re.
Merritt came to ThinkAdvisor Life/Health offices in New York for an interview Thursday. Here are five things Merritt about the future of MassMutual’s AI systems during the interview.
1. Life insurance is just the start.
MassMutual is working on applying similar AI-based analytical techniques to disability insurance and annuities, Merritt said.
MassMutual is also working on a risk scoring tool for people who want to apply for life insurance without providing samples of blood, saliva or other bodily fluids.
2. Insurance regulators seem to understand what MassMutual wants to do.
“They do have the right folks,” Merritt said.
Regulators want to put the consumer first, and they see that what MassMutual is doing will help consumers, Merritt said.
3. MassMutual anonymizes the data it uses.
Merritt said MassMutual respects the consumers’ privacy.
One benefit of the LifeScore360 approach is that the LifeScore360 system helps users see the factors that contributed to the score, Merritt said.
“Consumers want transparency,” Merritt said.
4. Risk scoring could help insurers improve how they tailor their products.
Today, Merritt said, insurers may reject some applicants outright because they do not know whether writing business for those applicants makes sense.
If insurers have better data analysis tools, they may be able to offer coverage to some of those people, Merritt said.
5. The LifeScore360 system has already produced interesting insights on common underwriting factors.
Researchers have published some academic studies suggesting that how body mass index correlates with people’s death rate changes as people grow older.
For insurers, one question has been whether that fluctuation exists at all.
Another question has been whether, even if the relationship between body mass index and mortality does change over time for members of the general application, the same change occurs for life insurance insureds.
In some cases, life insurance insureds are much different from members of the general population.
When MassMutual analysts looked at LifeScore360 data, they found that the relationship did change over time for life insurance insureds: for at least some older insureds, an increase in body mass index correlated with a lower death rate.
The effect for insureds was comparable to the effect for members of the general population, but shifted, Merritt said.
Showing that a rising body mass index may correlate with lower mortality for older insureds may be helpful in insurance underwriting because, up till now, insurance underwriters have typically used the same or similar body mass index guidelines for adult applicants of all ages, Merritt said.
— Read 1 in 6 Americans Are Hiding Secret Debt From Their Partner on ThinkAdvisor.