What is a CASoS?
- A complex, adaptive system of systems. They exhibit unexpected behaviors due to the complexity of their interconnections, they adapt, and they’re constructed of components sufficiently complex and difficult that each component can be considered a system in its own right. Systems involving multiple humans are generally CASoS.
Why are you working on CASoS rather than something else?
- All the world’s biggest issues (e.g. wars, energy, poverty, hunger, nukes, and so forth) are CASoS problems. We as a national security laboratory owe it to the world to take a shot at this.
- The most challenging problems of the coming decades will involve CASoS. In many ways this shift from more traditional engineering problems is due to increasingly pervasive and global scale of production and distribution networks in the areas of food, energy, water, labor, finance, etc. Expansion of these networks to the global scale along with increasing resource consumption has removed excess capacity from many interconnected distribution networks.
Why do you believe that CASoS systems are hard?
- They are big, complicated, counterintuitive, hard to measure, hard to modify because they adapt. They are generally very important, so they’re hard to experiment with (expense, difficulty in making changes to them, difficulty in measuring results, ethics). System and environment definition are often not clear-cut. The Adaptive nature of CASoS includes emergent and transient definition of systems and relationships between them and their environment.
Why do you believe that CASoS systems are sufficiently easy that you can do something with them? Why do you believe your process for dealing with CASoS systems is sufficient?
- We’ve been working with some (e.g. Pandemic Influenza). We’ve seen some interesting successes, frequently requiring a fresh look at the problem. It seems to take more than a first-order naive model of them. A systems view is generally necessary, and it requires thinking about multiple potential questions, multiple potential answers, and examining the interrelationships between them. Curiously, when we’ve constructed the "right" model, it often requires simpler models with easier-to-verify parameters than the typical models being constructed for them.
What is your CASoS process?
- We’re still working on it, but so far we’ve done well with broad, high-level thinking, combined with simulation that emphasizes the network nature of the system. We’re leveraging traditional systems engineering processes (requirements, conceptual design, detailed design, test, manufacture). We’re leveraging complexity theory, so we’re looking at ways of separating forests from trees, and ways of measuring when the complexity can be large. We understand that systems adapt, so we’re looking at solutions that change the system in ways that ensure that the solutions grow with the adaptations.
How is CASoS Engineering different than SoSE (system of systems engineering)?
- SoSE focuses on the structure of the system, and so SoSE doesn’t have to be complex or adaptive. Generally the concerns in SoSE are that the systems are evolving concurrently, are hard to manage, cannot be changed (unlike components which can be altered to make the system work). CASoS adds to this the notions that the system is complex (counterintuitive behavior a function of the interactions), and adaptive (the systems and the connections between systems change to adapt to changes in the environment, so solutions don’t keep working by themselves).
How do you deal with the fact that CASoS systems adapt? Doesn’t this mean that no solution will ever be adequate?
- It means that solutions have to also be adaptive, to change with the system as the system changes. It means that point solutions are probably not adequate, and that it’s a good idea to have multiple solutions in your pocket before you go trying to change a CASoS. Same thing as walking into a class of 3-year-old children. They adapt too. Don’t turn your back on them, have a wide ranging game plan, be prepared to adapt your solutions to the way the system works, and it can be a lot of fun.
Are you just making all this up?
- Well, yes and no. There isn’t a lot of theory out there, so we’re collecting what we can, trying it, adapting it to fit, and continuing forward. Such has been done before in any number of fields – when you’re trying something new, you sometimes need new tools.
- But, we have been working on some of these kinds of problems for some time. We know the feel of effective solutions, and know something about how to obtain them. We’re not going in empty-handed.
What are your tools for working on this?
- Our tools and approaches include complexity theory, systems engineering, system-of-system engineering; modeling and simulation; conceptual design; systems thinking; combinatorial consideration of interactions among system elements (random juxtaposition?); structured ways of thinking about the problem; broad unstructured ways of thinking about the problem; questioning the wisdom of previous attempts; paying attention to the underlying assumptions of previous solutions that might have at least partially worked.
Why do you believe various CASoS systems are comparable and deeply similar to each other?
- We’ve seen the same thing in classical systems engineering (compare electrical, hydraulic, spring-mass-dashpot, etc systems and you find identical underlying mathematics).
- Fractal behaviors are an important part of complex systems. Fractals are self-similar. We expect to see similarities across domains, as well as similarities between various levels of the same problem.
- In writing the CASoS roadmap, we discovered significant similarities across domains, both in their structure, in the kinds of models one might build, in the kinds of issues one might address in trying to solve something. Glass saw similarity between pandemic, wildfire, and the need to spread ideas. Both the technical solution and the political solution to the problem had structural similarities to putting out a fire (or starting one).
Isn’t it all just one big CASoS?
- Sure, constructed of systems which themselves are CASoS, and so forth. Great fleas have lesser fleas, upon their backs to bite ’em, and the little fleas have lesser fleas, and so ad infinitum.
If it is just all one big CASoS, where do you start?
- Almost anywhere that’s convenient. Initial thinking about any problem is cheap; it’s the details where things get expensive. So think hard about the issues, think hard about the potential solutions, think hard about how to change the world and achieve big wins — it’s a means of seeing many avenues to work on, and is relatively cheap compared to going down wrong, narrowly considered paths.
What are the ethics of changing the question you’re asking to match the answers you think you can find, rather than just solving the problems?
- The issue is in seeing the problems for what they really are. You might find that some of the biggest problems aren’t what you thought they were, but you have to see the big picture. Take any big problem, and you might conclude that it’s bigger than it really is (mountains out of mole hills), or that it’s only a problem because people insist that the solution benefit them in selfish ways.
How is this different from CAS work going on at places like Santa Fe Institute?
- Places like SFI are studying complex adaptive systems. We’re interested in engineering solutions for such systems. We are *very* interested in SFI’s results, but we’re also interested in solving problems. Recall that a stronglink/weaklink concept permits nuclear weapons to address a wide variety of potential threats in a very simple, robust way. Maybe we can find similar levers to address CASoS issues — it’s possible that the solutions can be very simple, but we have to think well about the problem to make sure we’re really solving real issues.
So why can’t we apply typical engineering approaches to these problems? What makes them so special/hard? Why do we need a big project in this area?
- Problems that arise from CASoS are, by definition, complex and are not well predicted or controlled by traditional engineering techniques. Here is a simple example problem to help make the case. Water security work at Sandia has resulted in algorithms and software to predict the next measurement of water quality based on several hundred recent measurements of that water quality. Water quality is a function of the characteristics of the water source and the treatment, amount of mixing with other waters and the flowpath it takes to reach the sensor. None of these things change drastically over time, treatment, mixing and aging are diffusive processes and the algorithms work quite well. Given that the stock market is also a time series, why do we still work at Sandia? Changes in the value of the stock market are one output of a globally connected CASoS and are considerably different than those that change water quality. Today, social-political-technical information that impacts the stock market and other complex systems moves at the speed of light (literally, across fiber optic cables) and combines with other information at multiple time scales in a non-diffusive manner and impacts the stock market (system of interest) in drastically different ways than we could possibly predict with traditional engineering tools.