Tuned In

Q&A with Michael Niemerg, FSA, MAAA, predictive modeling manager at Milliman

Photograph: Joel Maisonet

Who or what inspired you to study actuarial science?

Like many of us in the field, my path to actuarial science was a bit meandering. I initially wanted to be a marine, but I had to give up that aspiration due to medical issues. I started college majoring in business administration, but it didn’t take, and I was already plotting a way out after one semester.

After nearly giving up on math in high school due to a lack of interest, I ended up taking Calculus II. Even though it was hard, I enjoyed being challenged and decided I wanted to do something that was business-focused with an analytical focus as well. Luckily, someone a few years older than me from my small hometown (Altamont, Illinois) was also studying actuarial science at the same college. When I heard her describe it, it sounded exactly like what I was looking for. I switched my major immediately and haven’t looked back.

Video Exclusive: Michael Discusses Machine Learning, Innovation and Actuarial Science

What new and exciting developments are happening with machine learning (ML) and analytics?

There is a lot of cool stuff going on in natural language processing and reinforcement learning these days. In particular, more powerful and more massive language models are being released on a consistent basis. These models are capable of really sophisticated things like question answering, translation, story writing and much more.

Researchers also are developing reinforcement learning architectures that can play complex strategy games like StarCraft, where the state and action space is much, much larger than games that have historically served as benchmarks. In many cases, they are playing at human-level or superhuman performance.

It’s easy for actuaries to ignore these developments since they aren’t really that pertinent to what we do day-to-day, but I think it’s a good idea for us to keep a pulse on what’s happening in the larger community. You never know when new application areas might open up. In fact, natural language processing can be particularly useful to analyze electronic health records for underwriting purposes, and I’ve seen some preliminary research on using reinforcement learning for portfolio optimization.

What is your favorite part of your job?

I really enjoy most of the components of my job: data wrangling, thinking through model design problems, playing with new modeling techniques or data sets, and coding. However, at the end of the day, all of those things feel hollow if they aren’t employed to ultimately add value to the clients for whom I work. Nothing is more satisfying than coming up with a better solution and having other people see the value in it.

How important is innovation and collaboration in your field?

I think collaboration is table stakes in most professions these days, or you will quickly hit a ceiling in your professional career. We are social creatures, and we must operate in a social environment to thrive—no man is an island.

Innovation is a bit different, as its relative value has shifted at different points in my career depending on the particulars of what I was doing. But it has always mattered. Now that I exclusively work on predictive modeling, it matters more than ever. ML and statistics aren’t just rote processes that we push data through. Rather, they are pieces of a machine that we can swap out and recombine in a multitude of different ways to tackle problems in countless ways. The first step of being able to do that well is to have a large toolbox of tools and techniques. The innovation step really comes into play when you need to figure out the right way to put together these disparate elements in new and different ways to satisfy all of the real-world constraints of problems you encounter. Innovation can be very tough. Most truly innovative ideas fail—it’s the nature of the beast. Instead of letting it discourage you, you should embrace failure and channel it productively.

What about technology interests you?

At a macro level, all technologies enable us to do more with less. If you look at global extreme poverty rates, they have been on a downward trend for generations. That’s fueled by technology both directly and indirectly. The great amount of good this trend has bestowed upon the world is truly incredible to consider. That said, we always need to make sure we are good stewards of technology so it can continue to be a force for good and not be abused by bad actors.

I am particularly fascinated with how technology opens up the possibilities of what we can achieve. This has been particularly evident in the last decade or so in the ML world. More varied and larger data sets, new algorithms, more computing power and other forces have combined to make things that once seemed unachievable quite manageable. It’s a wild time with new technology being developed daily. The challenging part for me is having the self-discipline to not try out a cool new software package or technique just because, but instead be thoughtful about making sure it’s also a good candidate to solve the problem at hand.

What skills should actuaries strive to learn?

One set of skills that are undervalued or perhaps not even in people’s vocabularies are metacognitive skills: the ability to know yourself and reason about why you think the way you do. We often like to think of ourselves as logical creatures no matter how much science might tell us otherwise. Rather, I think what really happens is that emotion comes first, rationalization comes second and logic only enters into the equation to the extent that we invite it.

If you can be reasonably objective about your own strengths and weaknesses, you put yourself in a better position to take action—to either play to your strengths or figure out how to upskill where you’re relatively weak. Likewise, the more cognizant you are about why you have the emotional reactions that you do or why you take certain mental shortcuts, the more able you are to recognize flaws in your thinking and course correct.

To some extent, this sounds obvious, but I think all of us at times fail to reflect enough and do not recognize the cognitive traps we fall into simply because it’s so much easier not to consider it. While I strongly think these skills are important, I’m not sure what the best ways to develop these skills are. Probably the most important thing is to think they are valuable and continually reflect on them, and also to seek outside opinions so you don’t get trapped by biases hidden inside of your own thinking.

What do aspiring actuaries need to know?

When searching for that first job, do not worry too much about things like what specialty it is in or whether to go the insurance vs. consulting route. It’s more important to find a great supervisor and/or mentor. You really want to be a part of a team where you are challenged to keep improving yourself while feeling supported along the way.

Michael Niemerg, FSA, MAAA, is predictive modeling manager at Milliman.

Copyright © 2020 by the Society of Actuaries, Chicago, Illinois.