Complex Adaptive Systems

Updated: Aug 19

A complex adaptive system is a special class of complex system that has the capacity for adaptation. Like all complex systems, they consist of many elements, what are called agents, with these agents interacting in a nonlinear fashion creating a network of connections within which agents are acting and reacting to each other’s behavior. Through adaptation, agents have the capacity to synchronize their states or activities with other agents locally. Out of these local interactions, the system can self-organize with the emergence of globally coherent patterns of organization developing. There develops a complex dynamic between the bottom up motives of the individual agents and the top down macro scale system of organization, both of which are often driven by different agendas but are ultimately interdependent. There are many examples of complex adaptive systems from ant colonies to financial market to the human immune system, to democracies and all types of ecosystems.[1]


Complex adaptive systems are composed of agents. An agent is an actor that has the capacity to adapt their state, meaning that given some change within their environment they can in response adjust their own state. For example, if the agent is a player within a sports game, then if we throw a ball to the person he or she can catch that ball. They are able to do this because they have what is called a regulatory or control system. A control system of this kind consists of a sensor, controller, and an actuator. The person is using their optical sense to input information to their brain (the controller), that is then sending out a response to their mussels (the actuator), and through this process, they can adjust to generate the appropriate response to this change in their environment. And it is this same process through which a bird in an ecosystem or a trader within a market is receiving information, processing it and generating a response. Typically, these agents can only intercept and process a limited amount of local information, like a snail following a trail on the ground. It does not have a global vision of the whole terrain around it and it must simply respond to the local information available to it.


With this capacity of adaptation, agents have some degree of autonomy through which they can choose to synchronize or desynchronize their state with that of other agents within their local environment. We might also call this cooperation or competition. They typically do this based upon the costs and payoffs for choosing one of either option, and this cost-benefit ratio varies depending on the scenario, or what we might call the game they are engaged in with other agents. Some scenarios such as playing chess have very low incentives for cooperation while favoring competition. These are called zero-sum games. While other scenarios have a much lower cost and a higher payoff for cooperation; such as driving your car on the correct side of the road.[2] These different types of games create attractors that result in default positions for agents to cooperate or compete. Added to this are feedback loops, where what one agent does influences what another chooses to do. If you owned a certain stock and upon hearing some negative news about that company all of your fellow traders around you started selling it, this would create a positive feedback loop attracting you to also sell. And if you did, that would again amplify the positive feedback placing a stronger attraction on others to also do likewise. In such a fashion some phenomena can cascade through a population synchronizing their states rapidly.


This process previously described is a form of what is called self-organization. From the interaction of the individual agents arises some kind of global pattern which typically could not have been predicted from the behavior of the agents in isolation. For example, in the brain, consciousness is an emergent phenomenon which comes from the interaction between the brain cells. Thus, the global property of consciousness results from the aggregate behavior of individual elements. Within this macro-scale system that emerges control and regulation is typically distributed out.3 There is no master neuron or set of neurons that tell the whole brain what to do. No one is in control and no one in the system has complete information of it. This distributed nature of complex adaptive systems may make them very robust, where the system can adapt to some large disturbance. The internet might be an example of this. Dynamically updated routing tables keep track of how long it takes to send information along any path on the network. If there is a failure in one part of the network, packets are rerouted through another channel. Control over the flow of I.P. packets is distributed out over many different routers and service providers, with a large amount of redundancy making it robust to failure. But equally complex adaptive systems can self-organize into a critical state where feedback loops can work to amplify some small perturbation into a large systemic effect as witnessed during a financial crisis.


This emergent macro-scale system of organization then operates within some environment. Whether we are talking about a herd of animals within an ecosystem, the human body, a democracy or a corporation within a market, the whole macro system is periodically subject to perturbations and change within that environment. In order for it to optimize its state there must be some mechanism for performing selection upon the agents within the system. Those creatures within an ecosystem that can best respond to the environment are replicated. Those employees that have proven their value to the company will be promoted while others will be fired. Those products that best fulfill the demand are selected by the consumer, while others go by the wayside, the result being that the whole system evolves to exhibit more of the desired characteristics as they become more prevalent with the system.

Top-Down & Bottom-Up

In this way, this global pattern of organization will feed back to affect the agents on the local level, both enabling them and constraining them. It enables them as it is a mechanism for them to coordinate their activities and thus receive the benefits from forming part of a complex organization in the form of security, shared knowledge, technology, and so on. But it will also constrain them as following regulations and being subject to some form of selection is part of maintaining this global organization.[3] But of course agents have their own agendas that may or may not be aligned with those of the whole system, and this is where the real complexity comes into the dynamic, as there is now a core tension between the micro and macro levels. The system as a whole, that is, how it appears within its environment, will be primarily defined by how this core tension is resolved. That is to say, is the system driven by the interests of the agents at the expense of the whole? Or by the interest of the whole at the expense of the interests of the individuals? Or has it managed to find some resolution to this conflict? If we take an example of an economy, we can have a free market economy which is driven primarily by the interest of the agents in a bottom-up fashion, or we might have a communist economy driven by a top-down dynamic at the expense of individual motives, or we may have some economic system that manages to integrate the two.

1. What are complex adaptive systems? - Stockholm Resilience Centre. (2020). Retrieved 19 August 2020, from

2. Why Covid-19 and systemic risks are part of the hyper-connected world we live in - Stockholm Resilience Centre. (2020). Retrieved 19 August 2020, from

3. Top-down and bottom-up design | Wikiwand. (2020). Retrieved 19 August 2020, from,systems%20of%20the%20emergent%20system.

Systems Innovation

  • LinkedIn
  • YouTube
  • Twitter
  • Facebook