04Oct2012

What Is Complexity Science?

By Glenda Maki, Communications Program Coordinator, Society of Actuaries

Perhaps you’ve heard the name, but aren’t exactly sure what complexity science is. You’ve come to the right place! In this roundtable, six members of the SOA’s Forecasting & Futurism Section talk about complexity science and how it can benefit actuaries.

Participants:

Alberto Abalo, principal, Oliver Wyman
Donald Krouse, VP & appointed actuary, Transamerica
Clark Ramsey, VP & chief actuary, Employers Reassurance Corporation
David Snell, technology evangelist, RGA
Benjamin Wadsley, Transamerica

Ben Wolzenski, Actuarial Innovations LLC

How long has complexity science been around? Has it been referred to by other names?

Wadsley: To be fair, complexity and the study of complexity have been around a LONG time. That said, complexity science as we think of it today is fairly new. The goals of creating artificial intelligence can be traced to the beginning of the computer age. While many of the subtopics of complexity science emerged at different times, one of the landmarks was the invention of the genetic algorithm by John Holland in the 1960s at the University of Michigan. Some earlier computer scientists in the 1950s actually studied the idea of evolutionary systems with the goal of using it as an optimization tool for engineering problems.

Ramsey: There were many researchers in various fields dancing around the edges of complexity science long before the current name and techniques were adopted. One might, for example, interpret the invisible hand of Adam Smith as an early non-mathematical example of an emergent behavior, and similarly for Friedrich Hayek’s 1940s work on self-organization. Emergence of macro properties from simple micro-level rules and of self-organization without any external or central control, are properties of complex systems.

Snell: I believe the important thing for actuaries to know is that classic, deterministic means of solving problems have limitations in a world that has many interactions and no requirement to adhere to a theoretical model.

Abalo: I hate to begin my blogging career by referencing Wikipedia, but who can improve on this elegant chart ?

Is complexity science “science?”

Krouse: There are many definitions of “science.” From the perspective of a systematic approach to assembling knowledge and theorems that are testable and predictable, some aspects of complexity science may not neatly fit (after all, one of the objectives is to identify those areas that are not necessarily predictable). However, from the perspective of defining knowledge about natural systems, complexity science is certainly a science.

Abalo: I do not view “complexity science” as an individual field of study. Instead, I see it as a paradigm that touches on all areas of scientific inquiry, from mathematics and biology to economics and psychology—and, yes, even actuarial science. Where we typically study systems by breaking them into components, complexity makes us think of a system as we instinctually know it—greater and different from the sum of its parts. Ant colonies, locust swarms and human economies do not follow the rules we associate with the individuals that make up the system. In other words, we cannot define these systems in their entirety by defining how ants, locusts and humans act in isolation. Complexity differs from the current paradigm of scientific inquiry in that it employs inductive reasoning to investigate patterns of interaction and adaptation and concepts of emergence and self-organization in social systems.

Ramsey: Karl Popper taught that a theory is scientific only if it is “falsifiable,” or in other words, if that theory is false then it must be possible to so demonstrate experimentally or through observation. A theory that could not be falsified is in this sense not scientific. Many aspects of complexity science appear to be falsifiable and therefore to be science. Complexity science is a broad term with somewhat ill-defined boundaries, so it is certainly possible that not all of complexity science qualifies as science under this perspective.

Wadsley: According to Melanie Mitchell in her book, An Introduction to Genetic Algorithms, “Science arises from the very human desire to understand and control the world.” With the spirit of that definition of the motivation of science, complexity science fits into this nicely.

What are some of the key tools/techniques associated with complexity science?

Snell: Inductive reasoning takes many forms: genetic algorithms, cellular automata, serious games, Delphi studies, etc.

Wadsley: Some tools are being used today in the financial world (genetic algorithms and predictive modeling). Some have potential but it’s not been fully realized (network science, fractals, cellular automata and deterministic chaos). Others are heavily used, but are not considered by all to be included in complexity science (behavioral economics).

Wolzenski: I would cite (and strongly recommend) the landmark paper by Alan Mills, FSA, ND (©2010 Society of Actuaries): “Agent-based models are the heart of complexity science, and the four archetypal models are networks, cellular automata, artificial societies and serious games.”

Krouse: The Forecasting & Futurism Section has been busy compiling a list of resources and examples for each of these tools and techniques. These are available on the SOA website.

One can see how fractals can be used to measure coastlines or estimate carbon emissions in the rain forest. What are some of the applications?

Ramsey: The use of Mandelbrot’s multifractal models in modeling financial markets seems to hold promise in risk management and needs to be further explored. It offers another approach to the fat-tail problem, in which outliers occur more often than most models predict, and the tails are of course a key focus of risk management.

Abalo: In life insurance, advanced algorithms are replacing blood testing and medical exams. Today, those algorithms are created (or at least vetted) by a team of insurance professionals: underwriters, actuaries and administrators. Is it far-fetched to imagine a system that outsmarts these experts so that the algorithm eventually writes itself? Health care providers are already beginning to embrace the benefits of machine-based learning.

Wadsley: While there are several books that claim to use fractal for technical analysis, the real value may come from understanding other hard-to-measure “borders” such as value at risk (VaR) and conditional tail expectations (CTEs). Dynamical systems focus on the gross behavior of all solutions instead of analytically precise local behavior, which can be valuable in understanding tail risk.

Snell: Our DNA is folded into a fractal so that a 1.8 meter strand can fit neatly into 1/100th of a millimeter. Our lungs are fractals to maximize surface area for the given volume. Amazon.com sells many books on fractal analysis of the stock market. We may discover that a virus spreads in a fractal manner or that a beneficial bacterium may exhibit fractal tendencies. Perhaps a fractal is a more effective delivery mechanism for medicines.

Krouse: More generally we often find that actuaries are expected to be able to completely and accurately “predict” the future. Without a crystal ball, this is unrealistic. However, using some of the tools of complexity science, an actuary can define a set of conditions, assign probabilities, etc., and from these derive a likely outcome or range of outcomes.

As a simple example, we need look no further than the current low interest rate environment. By many “probabilistic” measures, current interest rates are highly unlikely—and yet they exist today. In hindsight, could some complexity science tools have identified these scenarios? Absolutely! Would people have reacted to them? Well, hindsight’s always 20/20.

Have traditional employers of actuaries been receptive to complexity science approaches to problems? Why/why not?

Snell: Property and casualty insurance companies are embracing complexity science techniques (especially predictive modeling) at a far more rapid rate than life insurance companies. Health insurance companies are now utilizing serious games to test therapies faster than possible (or allowable) with human patients.

Abalo: Why not ask if we, as insurance professionals, can cut through organizational red tape to convincingly demonstrate to our employers that novel approaches merit an investment in resources? The powers that be would likely be very receptive to approaches that improve the efficiency and accuracy of tools and processes we already use and potentially create competitive advantages by opening new areas of inquiry. I believe complexity science offers sufficient benefits to overcome the human tendency to simply do what has worked in the past.

Ramsey: Much of the work of actuaries involves projecting future cash flows and discounting them back—the details may differ from pricing an insurance product to setting a U.S. GAAP reserve to valuing a pension plan; but nonetheless, the idea that fundamentally we are projecting cash flows and discounting them back holds.

Traditional employers of actuaries will be receptive to complexity science approaches to problems when those approaches offer better ways to project cash flows and discount them back.

To date, I do not believe that actuaries have done much exploring of ways that complexity science techniques can improve our projections, but as the number of actuaries with an interest in, and familiarity with, complexity science increases, I believe that this will happen.

Krouse: I think the response has been varied. As with anything “new,” it does take time to become accepted. Having said that, we must also be cautious that complexity science tools are not viewed as a panacea. Like any actuarial or financial tool, there are certain applications where they are of best use.

Wadsley: As the field continues to be refined, I feel it will be increasingly used. “I expect that the children of 50 years from now will learn cellular automata before they learn algebra.”—Stephen Wolfram, 2006.

It looks like the Forecasting & Futurism Section first got involved with complexity science in 2010, and there have been some well-attended/highly rated sessions at SOA major meetings. What’s next in your plans? Or what would you like to see?

Krouse: We certainly have had some well-attended sessions in the past couple of years. We anticipate sponsoring our section’s first webcast later this year. Our longer-range objective is to raise awareness of the tools and techniques available. We plan to accomplish this initially through meeting sessions, webcasts and our section newsletter, but we are also working on including more complexity science topics in the SOA exam syllabus. Membership in the Forecasting & Futurism Section has grown significantly in the past two years. Obviously, actuaries are interested in these topics, and I encourage all members to become involved with the work that we are doing.

Ramsey: In addition to sessions at SOA meetings and webcasts, we will also continue to co-sponsor sessions and research, not just in complexity science but in other fields of interest to us as well, such as nontraditional forecasting techniques, demographics and futurism.

I would like to see a wider variety of tools for practicing actuaries to select from in order to better perform our roles. Some of these tools may come from complexity science approaches, but whether they arise from complexity science or elsewhere is less important than continuing to improve and expand our capabilities.

Snell: Genetic algorithms will become far more popular among actuaries in the coming year, as will serious games, cellular automata and behavioral science.

Wolzenski: Artificial society models can be used for sensitivity testing and forecasting. A background article on artificial society models was in the summer 2012 newsletter of the Forecasting & Futurism Section, and a further article will be in the winter 2012 newsletter.

Wadsley: Through my study of complexity science (genetic algorithms in particular), it appears that other professions have been much more successful at using these new techniques. I’d like to see us continue to use the lessons learned from those other professions to make our own successes and growth in the field.

Abalo: I understand our day jobs often demand 150 percent of our focus. With exams long behind us, it can be challenging to find time to explore new approaches, however interesting. Further, moving beyond satisfying personal interests and actually demonstrating use to an employer may seem impossible, if not laughable. I want actuaries interested in complexity to be able to avoid “starting from scratch” and turn to our section as a resource for continuing education and examples of how they can successfully implement these concepts in our corner of the universe. Examples like Ben Wadsley’s use of genetic algorithms in portfolio optimization and Ben Wolzenski’s explorations of how artificial societies can be used for forecasting insurance needs demonstrate that actuaries can—and should—use the tools emerging from complexity science. Stephen Hawking has said the 21st century will be the century of complexity. Let’s not miss out!

Sharing is caring.
  • Subscribe to our feed
  • Tweet about this post
  • Share this post on Facebook
  • Share this post on Google
  • Share this post on LinkedIn

Discussion

One response to "What Is Complexity Science?"

  • G Dale Meyer says:

    If one is not highly competent in computer science and mathematics is it possible to truly understand complexity science?

    I have a general notion of autocatalytic self-organizing systems. I was one of the pioneers in creating the relatively new field of entrepreneurship in academia. As this field has found its legitimacy all theory has been implanted by economics and economists. Within my limited abilities it seems to me that academic entrepreneurship can be better understood by theory built on complexity science. I am interested in working with a true expert in complexity science to bring my deep knowledge of entrepreneurship [as a serial entrepreneur plus a founder of the academic discipline of entrepreneurship] in partnership to create and publish the Complexity Science Theory of Entrepreneurship. This would be a first and lasting theory.

    G. Dale Meyer, Ph.D. gdalemeyer@gmail.com
    Distinguished Professor Emeritus of Strategy and Entrepreneurship
    University of Colorado – Boulder

    Write to me if interested and on the leading edge of complexity science. Thank you.

Leave a Comment