Seeking REAL Data
Q&A with technology evangelist Dave Snell April/May 2017Photograph: Whitney Curtis
Q: Why did you become an actuary?
A: I was an engineer in a research lab. One Thursday morning, I came to work and discovered that due to a defense cutback, our team was being laid off. At the time, I was young and still marketable. However, my colleague in the next office, despite being a world expert in fuel injection technology with 23 years of experience, had very limited options. There were only two firms in the world that could employ him in that capacity—and one had just laid him off. I witnessed the personal devastation of a very competent career engineer and decided never to become that person. That weekend, I accompanied my wife, who worked at an insurance company, to a work party. She introduced me to an actuary, who said, “If you like math, and you like money, this is a great career.” Mathematics had always been a passion for me, and with a two-month-old son, money sounded like a really good idea.
Q: Why did the actuarial profession attract you?
A: The actuarial profession offered the opportunity to work in a nontoxic, nonhazardous, not-politically-dependent environment with extremely bright colleagues and get paid very well for it. It satisfied my idealistic desires to help humanity by designing affordable and dependable products to protect families from dire financial need after the loss of a loved one. It also appealed to me that the exam process was so oblivious to gender and race. Career progress was highly related to exam progress, and it is really difficult to discriminate unfairly against a candidate number. Also, there appeared to be no ceiling for professional growth.
At the time I started working in this field, actuaries headed most life insurance companies in the country. The actuary was the only person able to master commutation functions to do the complicated mathematical forecasting needed for pricing, valuation and other “quant” tasks. But the actuary also was able to cross over into other business functions. He or she could converse with an underwriter about the longevity significance of a complicated disease, meet with the investment department about the beta of a stock or bond, talk to the information technology team about storage bytes, and chat with the marketing folks about consumer needs and preferences. Only the actuary spoke all of these languages and had empathy for the many areas of an insurance company. Strategic leadership was a natural outgrowth of that ability to understand both numbers and people—and to convey goals and results in a lucid manner.
Unfortunately, I fear we have lost much ground since the ubiquity of electronic spreadsheets that eliminated the need for commutation functions and the invasion by other quants (MBAs, CFAs, CPAs, etc.) with similar computing tools but better communication skills. There is an old Chinese saying—“I wish my enemy 50 years of prosperity”—that seems to apply here. The idea behind the saying is that a long period of prosperity can erode our initiative and our willingness to adapt to changing times. Actuaries, as a profession, got complacent. Now, on some fronts, such as predictive analytics, we must play catch-up.
Q: How did your professional experience lead you to a career that is somewhat less traditional?
A: Back in the late 1980s, I had the unusual (for an actuary) opportunity to move into the sales field. I lived for three years on commissions from managing 1,200 brokers in the San Francisco area. My initial assumption was that my vast product knowledge would make me a brokerage magnet. Unbeknownst to me, the term “actuary” was synonymous with “deal killer” in the sales environment. I had to become a chartered life underwriter (CLU) and chartered financial consultant (ChFC) to get some street cred. I also started teaching agents how to pass their life underwriter training council (LUTC) exams.
By the third year, they elected me president of the Life Underwriters Association—but by then I realized my strengths were not in direct sales. Often, I would convince clients to stay with their own plans rather than buy the one the broker and I were selling. I returned to a home office environment as a “respectable failure,” but I was far savvier than when I had left it. The education was expensive, but it served me well afterward because I had learned empathy for the sales process that I could not have gained in a comfortable salaried position. In a sense, this was my introduction to behavioral economics, which play an important role in effective predictive analytics.
Q: How did you segue into work in predictive analytics?
A: In 1982, a good friend (who later became our CEO) gave me a copy of Gödel, Escher, Bach: An Eternal Golden Braid, by Douglas Hofstadter. I became fascinated by this new area called artificial intelligence (AI). Later, he had me develop an AI expert system for life insurance underwriting. RGA was a pioneer in the development of underwriting expert systems. It has been translated into several languages, including Chinese, and currently is being used in many countries. More recently, I was co-inventor on a patent to combine machine intelligence with human intelligence.
Q: What brings you the most joy in your current position?
A: I love working with, inspiring and learning from the bright young minds around me. We have a lot of Ph.D.s in our group. Relative to any of them, I am academically challenged; but we seem to have a synergy that produces amazing results. I’ve been blessed to have the pleasure and pride of teaching individuals to do something (or to try something) and then watch them do it better than I could have ever imagined myself.
Q: With regard to predictive analytics, what skills positioned you for work in this area?
A: I had a nonstandard experience set for the predictive analytics area. Fortunately, our company recognizes that a wide mix of people with complementary skills is needed to make predictive analytics projects successful. My strong computer programming background, which spans four decades and 30 programming languages, was a big asset because so much of predictive analytics involves harvesting and preparing the “tidy data” before you can use it. The actuarial studies provided me with a decent background in statistics, and I strengthened my knowledge where necessary through books, online courses and interactions with kindred spirits. The Society of Actuaries (SOA) Predictive Analytics and Futurism (PAF) Section is an excellent place to shortcut this process.
Q: What skills do actuaries bring to analytics that other professionals may not?
A: In addition to the stats and IT knowledge, a data scientist for financial risk needs subject-matter expertise. Other quants may actually be stronger than actuaries on the first two items, but we shine on this third one. Our PAF newsletter has several articles with examples of how important subject-matter expertise is to the successful creation and use of predictive models.
One glaring example of how important it is to understand the assumptions of the algorithms utilized is the “dead salmon syndrome.” Thousands of functional magnetic resonance imaging (fMRI) studies have been used to validate new medications and medical procedures, and a few years ago it was discovered that the underlying algorithm to interpret these fMRI results had a flaw that gave a lot of false positives. It was confirmed after a researcher hooked the electrical leads to a dead salmon and got results supporting the efficacy of an experimental procedure. Keep in mind that the subject was a salmon—not a human brain—and it was dead! The actuary brings a strong background of experience and knowledge that permits a more informed judgment of results from these models, or of the validity and applicability of underlying assumptions.
Q: What kinds of challenges are you solving using data analytics? How are they different from the issues you would address in a more mainstream role?
A: Many data scientists would say that getting enough (and relevant) data is most important. I think that is very important, but critically looking at your data and your algorithms from a more holistic perspective also is a huge challenge.
In our company, we like to say that we seek data that is REAL: relevant, ethical, affordable and legal. It is obvious that relevancy is important. If you throw a bunch of confetti up into the air and then look at where it lands on the sidewalk, you will likely discover that you can neatly draw a chalk line around some clusters. These represent correlation but probably no meaningful causation. Lots of cross validation techniques exist, but I think external validation is also important. This is especially true when using techniques such as deep neural networks, where it is difficult, at best, to trace an answer back to the assumptions and rules employed.
We owe our internal decision makers (such as senior management) and the public the due diligence required to justify their trust in us. A more mainstream role might utilize a known formula where you offer others the opportunity to see all the steps involved to the level of detail they wish. Again, the opaqueness of some predictive analytics techniques is a challenge, but it is also an opportunity for actuaries to show the value they add through their subject-matter expertise.
Another challenge is that sophisticated models can obscure assumptions or rules that are not ethical or legal. If a multivariate model turns out to incorporate ZIP codes or a proxy for them, it may be guilty of racial profiling. We need to be the watchdogs of our own industry so others do not feel the need to impose those controls upon us.
Q: Where do you see opportunities for actuaries in the predictive analytics arena?
A: The obvious areas are in our own industry: life and health insurance, general insurance, valuations and so on. But those are just the tip of the iceberg. The SOA Cultivate Opportunities Team is promoting actuaries in predictive analytics roles for investments, enterprise risk management, manufacturing, software development and many other business areas. Some actuaries might be the ones building these models, others may be using them, and still others may be managing the associated risk departments.
Q: How do you see the role of predictive analytics changing in the next five to 10 years? Where will actuaries fit into the equation?
A: Many companies are jumping into predictive analytics. As data becomes more readily available and computers become more powerful, it is creating a near-term boom in the demand for predictive analytics and those who can use it. My personal feeling is that this will lead to some very good implementations and some that will be more challenging. Actuaries can help ensure that the good results are based upon REAL data and REAL algorithms. We have a professional responsibility to detect, change and, where possible, prevent the misuses of predictive analytics.
Q: What are some of your best professional experiences/memories as an actuary that may inspire others to explore different actuarial paths?
A: Several years ago, while installing some actuarial software in one of our offices in Asia, I was alone all night in the office and found myself unable to properly use the telephone, copier, fax machine and even the PCs (I made some PCs inoperable by misunderstanding the Chinese character error messages). I felt so angry at my own ignorance that when I returned to the United States, I began learning Mandarin Chinese. This was a personal hobby, but a few years later, when we needed someone to build and head an Asia-Pacific Technology team, I was chosen partly for my (admittedly poor) Chinese character fluency. My family and I had the pleasure of residing in Sydney for three years and of traveling throughout much of Asia. It was a mind-opening experience that gave us an appreciation for cultures far different (yet amazing!) from our own. You never know when some optional learning beyond the SOA study notes will pay large dividends!
Q: What is your dream job?
A: My dream job is one where I feel needed, challenged and appreciated. Almost every job I have ever had has become my dream job. Too many people think that they are not empowered to enhance their own jobs. That’s a very narrow-minded viewpoint. If you find that you do not like the job you are in, I offer three choices:
- Learn to like it.
- Change it to your liking.
- Leave it for one you can like or change to your liking.
Nobody should waste precious years in a job that is not his or her dream job.