Complex Adaptive Systems: Primer 1/N

Over the past year, I have dedicated a large portion of my research to the study of complex systems. This started out as me being bored in an econometrics course and reading random Wikipedia pages during class without any specific goal in mind.

Protip 1: If you’re finding yourself bored in class and are desperate for an escape, use class time to teach yourself about things that interest you; this is a classic case of using opportunity costs wisely.

Protip 2: Diversify your portfolio; sometimes the best ideas come from reading things unrelated to your research interests.

Since then, my knowledge and understanding of complex systems have come into fruition to something greater and will probably form the basis of my dissertation. Without further ado, let’s start our journey into complexity:

Systems are considered complex if they meet the following criteria:

  1.  The system is composed of interacting agents.
  2.  The system exhibits emergent properties.

The second criterion is of utmost importance. If you think of a system of gears (the gears being interacting agents), like a manual transmission, there is nothing unexpected that can happen. In my Toyota Matrix, I go through the same procedure every time I find myself stopped at a stop sign: 1) Disengage clutch, 2) Shift into first, 3) Let out clutch until I can feel the friction point, 4) Engage accelerator gently, 5) When car begins moving, remove my foot from the clutch. If I have done everything correctly, I will be on my way towards my destination. If I don’t, the engine will die and I will have to start again. This is a binary outcome. The transmission only allows for these two outcomes; even though, my transmission allows for eight different parameters: six gears, reverse, and neutral. No other outcome or pattern can emerge from this set of parameter and inputs (throttle and clutch engagement); the gears cannot create new combinations of movement, nor can they decide to go in any new direction. The transmission is closed system. It is not complex because of the inability for it to create new modes or patterns of interaction between the agents comprising the system.

This is a far cry from social systems which allow for pattern formation and new types of interactions. Patterns in our systems exist everywhere: The American system of government is pattern formed through over 200 hundred years of social interactions piled on top of the initial foray in 1776. This and all other forms of governance did not exist prior to the system, but are instead outcomes of the system. Importantly, forms of governance, from that of the family to corporations to governments to international organizations are the result of the interactions of individuals. Every so often, we create new modes of interacting: language, Morse code, Bell’s telephone, radio signals, television, cellphones, internet, email, texting, social media, Snap Chat, etc.

Talking about social systems allows us to introduce an additional criterion:

3. The agents can change based on prior and current conditions/parameters endogenous to the system; at this point the system can be classified as a complex adaptive system.

Think of how individuals and groups adapt to changing conditions in a social system. In the 1980s, laws were passed requiring that only certified electricians be allowed to work on electrical wiring in residential/commercial properties. This subsequently increased the cost of hiring an electrician; in response, homeowners were more likely to work on their own residential electrical systems. This was the first adaptation. The 1980s saw a rise in residential electrocutions as under-qualified individuals attempted to repair their homes. A second adaptation would have been one of homeowners realizing the dangers and then preferring to hire qualified electricians instead of risking injury. This is just one slice of adaptations; it doesn’t capture all the schools who expanded their electrician programs, those who lost work because they lacked the new qualifications, and etc.

Interactions breed consequences (not necessarily a negative term) and adaptations.


The Efficacy of Groups, Group Selection, and an Ecology of Plans

Richard Wagner posits that the macroeconomy is made up of an ecology of plans; I am sympathetic to his views because this allows for a framework of the economy to be seen as more than just the sum of its parts. This is because macroeconomic action is not just an aggregate of microeconomic action. The exception to this is at the very first encounters where the macro level interactions have yet to be formed (there are no institutions, formal or informal, that dictate behavior). After these are established, the micro transactions rely on the macro economy to enable them while the macro economy can only be perpetuated by the continuance of micro-level human action. One cannot exist without the other once the cycle has been initiated. Though, it is very possible that either one of these may wane in presence of the other.

The mediator between micro and macro action is that of the meso-level. Agents form themselves into groups; in fact, many agents will self-select or be selected into several groups. Families are one such type of these groups, while political parties, friendships, civic organizations, and religious congregations are all examples of groups in which any one agent can simultaneously take part. In both the public and private sphere, what causes these groups to survive throughout more than just one generation? What about an even smaller time scale, like more than a few meetings? What about Black Swan groups like the Bolsheviks? What are the behavioral mechanisms that ensure their continued survival? How does this compare to those who do not propagate for more than one life cycle?

I do believe that in the case of many of these sets, there exists a form of group selection similar to that in the evo-bio literature. Groups evolve a specific set of geno/phenotypic traits that occur at the group level instead of at the individual level. There is some argument in the evo-bio literature, but given that social systems can exhibit increasing returns to scale because of institutions or technology, I intend to sidestep their disagreements until another time. This means that a social system (a collection of groups into a very large group) is able to evolve certain traits that are different from those of another social system. I’ve argued in short essays that these traits may be readily copied by another group because knowledge is non-excludable and nonrival in nature. It has been pointed out to me that this means very little because it depends not on transference of knowledge but on the use of knowledge. I can’t agree more given that most people today have access to the entirety of human knowledge via a device in their pockets, but instead of using it for the advancement of our species, many play video games or feed a dopamine addiction. (This is pot calling the kettle black; I guilty of both of these.)

What I propose is an extension of Dr. Wagner’s hypothesis: the ecology of plans matters at the meso-level as well. Groups have plans. They are a way to lower the transaction costs of many people into a singular goal. Some of these groups seek domination of an entire economic system, others simply want to enjoy the fellowship of their members. I think I’ve mentioned Ostrom’s rules on common pool resources; these extend to efficacy of groups. In future posts, I hope to work out some agent based modelling of this.

Sidebar for myself: Demand will eventually create a supply through a variety of mechanisms and processes that necessitate the actions of entrepreneurial agents, but the converse is not true.

Ostrom, CPRs, and the Efficacy of Groups

Elinor Ostrom’s work in her study of common pool resources is paramount to my ideas. Ostrom defined the conditions in the form of eight core design principles:

  1. Clearly defined boundaries
  2. Proportional equivalence between benefits and costs
  3. Collective choice arrangements
  4. Monitoring
  5. Graduated sanctions
  6. Fast and fair conflict resolution
  7. Local autonomy
  8. Appropriate relations with other tiers of rule-making authority

These design principles are what is necessary for group survival and success to take place. Ostrom also defines success, for institutions: institutions are successful when they enable agents within the system to engage in productive outcomes despite the ever-present temptations of shirking and free-riding. I think this focus on “productive outcomes” could lay the basis of future system effectiveness/ efficiency, but that still leaves and immense amount of gray area to be resolved.

Using the CPR framework, the Soviet Union was destined to fail because its scores on conditions 2, 3, 6,7, and 8 were subpar. I would further claim that the USSR was doomed to failure (by productive outcomes) because it failed to meet certain criteria applied to socio-ecological systems. Namely, it failed to create a coherent system that allowed for interaction in a resilient and sustained manner. Furthermore, it failed to allow for a dynamic economy that allowed for continuous adaptation. Central planning and mono-centricity are anathema to adaptivity at the micro and meso levels.

[1] Wilson, Ostrom, and Cox. “Generalizing the core design principles for the efficacy of groups.” Journal of Economic Behavior & Organization, 90S (2013) S21–S32. 2012.

[2] Ostrom, Elinor. Governing the Commons. Cambridge, 2018.

Comparative economic systems, thoughts

As I alluded to in an earlier post, no economic system is purely capitalistic or socialist– there are an infinite number of potential combinations (this is analogous to sociology’s proposal that no person is purely masculine or feminine). This is not to say that one combination can clearly be superior to any other; though, at any given time, there might be one system combination may produce more output than another. Furthermore, there may be other metrics than the dichotomy that I have just posited. Surely, there are numerous types of economic organization that we have long forgotten or have yet to discover.

Some systems, even a few currently in existence, fail to expand past their current boundaries because of scalability issues. We know that issues with moral hazard and adverse selection often lead to the downfall of interpersonal and interfirm relationships; at the very least, they can put considerable strain on these same relationships. In the market, insurance companies and banks have to find contractual work-arounds and screening devices to limit these problems. Depending on the constraints and the desired outcomes, not all economic systems are created equally. Some systems may perform better for a certain task, at a certain time.

This poses a number of questions:

  • What are these outcomes?
  • What metrics do we use to measure the economy?
  • Are these metrics correct?
  • What is the goal of the system?
  • What is the paradigm that governs the system?
  • What are the base ideas of the system?
  • How do these ideas change the system?
  • What are the marginal changes of black swans?
  • Are averages overrated? (I think so)
  • What is inequality, and by what metric do we determine inequality?
  • What are the feedback loops in each economic system?
  • How do stocks and flows change these?
  • Which feedback loops have increasing returns to scale vs those that have diminishing return to scale?
  • What about returns to scope?
  • How can open ended evolution systems help us find the answers? (In a way that DSGE is unable to do?)
  • What are the ultimate ends of the society?
  • Was George Bernard Shaw correct when he stated that socialists were just communists without courage?

Upcoming, a look at the Soviet economy via Janos Kornai.