End Paper

Simplicity Through Complexity

Simplicity Through Complexity

By Sam Gutterman

I rarely see things in black-and-white terms. Rather, I always seem to focus on the shades of gray—advantages and costs, weighing a set of considerations.

Maybe I got into this habit while studying for actuarial essay exams. I can easily complexify any issue or problem, considering all relevant nuances (and some not-so-relevant). Sometimes I’m envious of the person who can make a decision based on what, to me, is incomplete information, especially when there is tail risk. At other times I am suspicious of any such conclusion.

Is this a personality flaw, or the proper way to think and make—or not make—decisions? To try to figure this out, I have to separate my answer into two parts: the thinking or analytical process and the communication of conclusions.

First, to best approach an issue or risk, I must understand it. Simple statistical extrapolation is to me unacceptable, except as a last resort. My preferred process is breaking down an issue into its component parts, together with assessment of (1) drivers, (2) interactions and entanglements, and (3) consequences.

And of course, in some cases the process followed can be more important than the end goal.

Approaches I have used range from the hard science of predictive analytics (parts of which I used to call multivariate analysis) to the softer science of focus groups or listening to others—the statistically advanced, to a search to better understand behaviors and beliefs. I try to use both in my search of the “truth.”

If the problem I am addressing involves the development of a projection, I need some healthy skepticism about the relationships involved, because recognition of cause-and-effect relationships among data or drivers is the gold standard. For instance, an effective study of mortality requires that I understand the reasons for trends and risks involved in the mix of direct causes of death, drivers of those causes, and their timing.

However, I also attempt to be careful about possible over-specification. The first time I looked at an analysis using a neural network, I got an answer, but it really only spit back the raw data that was used and provided me little insight or comfort that the results could be used for something other than predicting what already happened. Enhanced model governance is also needed and may be made more difficult if a complex model is used.

Striving for more knowledge is important, but understanding is not always enough. Better is to ensure that the question being asked is the most appropriate one. My son Jordan is particularly good at synthesizing the key feature of a mass of information, particularly if it relates to a political issue or development. This means figuring out the most significant elements that can lead to a practical and effective solution, recommendation, or insight. And the degree of “accuracy” or refinement needed will differ by project or topic. For many decisions, 80% right or 80% of an explanation is sufficient, while in other cases greater depth is needed.

After synthesis comes effective communication—in simple terms or achievable goals. But I don’t mean “simple” in a paternalistic way. Rather, my message has to be appropriate to the user, not the messenger. I must know the decision-making process followed by my users and what their expected takeaways are.

Today’s decision-makers often don’t have the time or patience to dig into or understand a complex model or its logic. When I oversaw a management reporting unit, I continuously challenged my staff to engage their audience: What did they expect the users of the information to do with the message?

Once many years ago I got into an argument with Hans Bühlmann, a famous Swiss actuary. He claimed that anything of actuarial importance needs to be expressed in terms of an equation, while I maintained that to be worthwhile, a message has to be expressed clearly in words that the user is willing to take the time to understand and develop consequential action about.

We agreed to disagree.

Education is often presented as a solution, and effective education is certainly useful—but it should not consist of a complex message. Use a graph rather than a table of numbers. Speech coaches told me long ago: Make your point, explain your point, and finish by making your point. The message needs to satisfy the “Keep It Simple, Stupid” approach—appropriate to the intended user—even though it is developed through complexity.

SAM GUTTERMAN, a member of the Academy and a fellow of the Society of Actuaries and the Casualty Actuarial Society, retired after many years as a director and consulting actuary with PricewaterhouseCoopers LLP in Chicago.

Next article The Election
Previous article Bugging Out—Drug-resistant pathogens, antibiotics … and the race against the clock

Related posts