An Overview of Complex Adaptive Systems

Following from the Presenting Complexity Theory article, Complex Adaptive Systems facilitate a pragmatic framework for complexity thinking with regards to tackling the uncertain nature of social systems and have emerged as an impressive framework for exploring and managing the behaviour patterns of adaptive agents, as well as other phenomena.

A Complex Adaptive System (CAS) refers to a system that is characterised by having a substantial number of agents that have corelated interactions with one another. CAS display self-organisation, in that control tends to be highly dispersed and decentralized (1). Self-organisation and nonlinear behaviour create emergent, global properties for the whole system (1). The overall coherent behaviour of the system is the result of a huge number of decisions made every moment by many individual agents. An example of a CAS is forests, organizations, stock markets and communities. We cannot predict the behaviour of a whole CAS from the basis of its constituent agents (due to nonlinearity) (2), as it is possible for a small change in one variable to shift the entire system into a very different state (such as consumers in a market) (3). A CAS has a memory and time base. It is a path dependent system that is anchored by its initial conditions, so that identical forces acted onto seemingly similar systems might react differently depending on their history (2). CAS have a degree of resilience to external forces depending on factors such as the diversity of its agents, the quality of network links and its proximity to any critical thresholds (2).

Agents

Agents are any constituents of the system that affect the system (4). They are commonly referred to as adaptive agents, highlighting their interdependent relationship of change where the subject discretely affects the environment and is also shaped by it (5). While it is common to use the term particle to describe components of a system, the term agent implies the autonomy or intent of components (1). Epstein’s (2009, p 5-6) formulation of agents being heterogenous implies that the agents have a range of values for their attributes, rather than having a range of attributes (1). Alternatively, we could describe them as being “self-similar” which recognises that agents in a CAS can be “different—but not too different — in terms of the rules and attributes that relate to the emergent property in question” (1).

Feedback

We can describe feedback as a process in which part of the output is cycled back or fed-back into the input (or causes) forming a circular loop mechanism in the system (6). Feedback can also be defined as the “outputs a system at time t affect the inputs of that system at time t+1” (1). Put simply, the outcome of interactions between agents has the capacity to influence some future interactions. This influence depicts the feedback in the CAS. This action/response relationship produces similar patterns in light of changing components. But it does not produce persistent patterns, in the sense that if the outcome of specific interactions initiates a particular interaction, similar outcomes will not always initiate that particular interaction every time (1). Multiple causalities are often sustained through circular feedback mechanisms (6).

There are two fundamental feedback patterns: Positive feedback: which increases the deviation from a stable state (change expanding) and negative feedback (change dampening) which does the opposite (7). Thomas Schelling’s “go-with-the winner strategy” describes how individual’s decisions are determined by the behaviour of others, even if those behaviours are against moral or social rules (e.g., crossing during a red light when others do it). Feedback structures in complex systems are influenced by notions such as “leadership, cooperativity, economic and political stability and even fashion trends”. For this reason, global phenomena in complex systems derives from its ability for self-organisation.

Self-Organisation

Self-organisation in a CAS means that no component (or any entity outside the system) can have direct or exclusive control over its individual and collective patterns (3). The parts in the system are not centrally controlled, they are controlled locally through physical and cognitive dimensions (3). Self-organising systems are characterised by bottom-up interactions and display coherence through out the system (7).

The “agent follows its own local rules, and uses its own attributes in applying those rules” (1). That is, individual agents act according to limited knowledge and local rules and provide influential feedback to other agents so that they may produce corelated behaviours (1). In this way, the system regulates its own flow of “information, energy and matter to produce new forms of particular order” (Bak, 1996; Kauffman, 1995) (7). The degree of variety and cooperativity (a type of behaviour in which various independent agents act collectively or synchronized) of a system’s constituents can affect self-organisation and thus emergent phenomena (1, 6). A good analogy to illustrate this is through a simple model of traffic flow, where each agent is a car moving on a highway. Under normal conditions, each agent is governed by two rules: To slow down when the car ahead of it is too close and to speed up if the car ahead is too far. Under these conditions, agents will slow down or speed up depending on the behaviour of the agent in front of it, creating a wave-like pattern to emerge. This patterned emergent phenomenon will occur even if there is a slight variation to activate each rule (heterogenous attribute values). But if some agents have heterogenous attributes, such as rules that allow them to stop, crash or drive off the road, then this chaotic behaviour would be too disruptive for any emergent patterns to form (1). Self-organisation is the process by which emergence occurs (8).

Emergence

The emergent properties or phenomena of a CAS are observable characteristic of a system that are at a different scale than the parts of the system (1, 9). Globally coherent and observable states or patterns of a system are determined through previous states of the system and the underlying local interactions of its constituents as a result of self organisation mechanisms. (6, 10). “Global order built out of local interactions” (1). Because emergent phenomena arises due to underlying dynamic interactions, CAS are irreducible; these high level, observable phenomena, cannot be reduced to lower level intricate states (3).

Paley (2010), believes the global state or emerging phenomena of the system is not intended by the agents: “no intention… plans or goals to produce it” which is further supported by Jones’ (2007) consideration that the properties of emergence may be something that individual agents might not possess themselves (9). It is possible that a social system made up of different agents can produce the same phenomena and this emphasizes the self-organising aspect of CAS: order occurs because of individual units and their environment conforming to local stimulus-response interactions (10).

In describing the functioning of complex systems, Inayatullah (1998) created an iceberg metaphor. The visible part of the iceberg coincides with the descriptive properties of the phenomenon, which is generally presented as the name characterizing the challenge, such as stigma, adherence to medication or disclosure. The submerged aspect of the iceberg coincides with the underlying interactions that influence and sustain the visible descriptor of the phenomenon (11). Self regulation of the system is generated through feedback between and within the visible characteristics or descriptor of the phenomenon and the distinct underlying. The underlying properties are made up of a variety of interactions between agents, feedback mechanisms that link and influence different agents within the system and linear and non-linear interactions (11, 12).

Unique forms of localised order (through self-organisation) lead to the emergence of system-wide, observable phenomena. The properties of this global phenomena can then feed back down to the level of the components and impose forces on their behaviour and interactions (such as constraining their choices and potential feedback dynamics, making self-organising systems self sustaining over time (Pincus & Metten, 2010) (7). These systems are thus bound together from multiple directions: nonlinear feedback interactions among system components in a horizontal manner and vertical bi-directional feedbacks between global and local scales (7). These systems also adapt or evolve depending on internal demands among system components or demands from the broader systems in which they are embedded (7).

Management of complex systems requires shifting the underlying feedback dynamics between the agents resulting in changes in the emergent phenomena (12). From this point of view, the AIDS and HIV epidemic can be understood as an eventuality due to the emergence from the feedback dynamics between the underlying constituents of the human-virus-environment interaction, which sustain the ‘HIV/AIDS phenomenon’ (12). There is much potential in tackling the discrete, underlying properties of the complex system rather than focusing solely on surface level characteristics. In order to implement this, insight into how underlying feedback dynamics affect observable characteristics and consequently emergent outcomes is necessary (12). One mediator of dynamic feedback processes that can be the focus of a management strategy in complex systems are strange attractors (12).

Strange Attractors

Self-organisation (and emergent pattern formation) in a complex system does not produce completely random nor predictable outcomes. The properties or patterns that emerge are limited to the “permeable boundaries” which the interactions find themselves in (11). The specific behaviours of agents in a complex system are unforeseeable, but they are roughly confined to a limited range or subset of actions and display observable preferences with regards to the complex system they most frequently interact with (11, 13). We can see this range of behaviours in different social systems that have distinct patterns of clustering and coherent, synchronised states. Strange attractors create “bounded instability” for complex, nonlinear systems. That is to say, they create parameters from which the system will be confined in and agents in the systems will naturally gravitate towards unless disturbed.

Strange attractors for any system can be represented through a phase space, which is a mathematical model that represents every possible state that a system can take. Each possible state of the system corresponds to one distinct point in the phase space. If we plot the agents’ interactions on the phase space, we will see that they do not pass through every state, rather, they are limited to a subset of all the possible states. In this way, strange attractors can depict patterns of behaviour in a phase space, which are never replicated, yet they always show similarity (13, 14). An example of this are the patterns in which HIV positive people take biomedical services and technology in a particular landscape. In SA up to 25% of people living with HIV are inconsistent with their ARV adherence and clinic follow ups and it is common to find people mixing traditional medication with biomedical medication (13).

Strange attractors are discrete influencers of emergent phenomenon and “represent a loci of change opportunities when managing complex systems” (12). When interpreted as a mix of patterned and particular emergent symbols that influence social practices produced through localised sense-making – with practices creating a positive feedback loop that influences the localised significance of attractors. Side note: Interpreting attractors as states of least resistance towards maintaining and restoring equilibrium, provides us with a pragmatic method to apply this knowledge. Disturbing attractors (and the underlying feedback between agents) can shift the system from the prevailing equilibrium and subsequently allowing the potential to interrupt the system with “non-linear, emergent change opportunities” (15). Disturbing and altering these attractors could produce alternative pathways to bio-social practices that hinder the viral load transmission in communities (16).

Resilience

The resilience of a CAS can be defined as the nature of flexibility of the system, or in other words, the system’s ability to respond to external shocks “by either becoming rigid or robust, or flexible and fluid without becoming stuck or falling apart respectively” (7).

In complex, dynamic networks dominated by strange attractors, a HIV positive person who binge drinks and doesn’t adhere to their medication reduces the options for personal level medical

responses by acquiring a combination of liver failure and MDR. By the virus being resistance to medication because of rapid mutation establishes conditions for the growth of new HIV strains at multiple levels, causing a public health crisis (13).

The authors stress “that many of the rogue social systems in the HIV landscape in South Africa are complex adaptive systems within which the strange and periodic attractors represent properties of the system that act like magnetic forces, constantly drawing the systems in patterned trajectories towards an equilibrium point which will jeopardize the goal of achieving Vision 90-90-90.” (13).

Conclusion

Complex Adaptive Systems provides us with a lens through which we can better understand multi-causal dynamics within the epidemiological landscape of the HIV/AIDS. They are a guide towards appropriate ways to manage, plan, design, implement and evaluate with respect to the degree of complexity of the issue in question. With the growing consensus of the complex reality in reaching Vision 90-90-90 and ending AIDS by 2030, complexity theory should inform the array of organising principles that would guide a comprehensive intervention instead of “dogmatically” applying linear approaches to human interactions.

---

Work Cited

1. Carmichael T, Hadžikadić M. The fundamentals of complex adaptive systems. Complex adaptive systems: Springer; 2019. p. 1-16.

2. Van Beurden EK, Kia AM, Zask A, Dietrich U, Rose L. Making sense in a complex landscape: how the Cynefin Framework from Complex Adaptive Systems Theory can inform health promotion practice. Health promotion international. 2013;28(1):73-83.

3. Turner JR, Baker RM. Complexity theory: An overview with potential applications for the social sciences. Systems. 2019;7(1):4.

4. Burman C, Mamabolo R, Aphane M, Lebese P, Delobelle P. The South African developmental landscape: restricted potentials or expansive, complex adaptive opportunities? TD: The Journal for Transdisciplinary Research in Southern Africa. 2013;9(1):17-37.

5. Burman CJ, Aphane M, Mtapuri O, Delobelle P. Expanding the prevention armamentarium portfolio: A framework for promoting HIV-Conversant Communities within a complex, adaptive epidemiological landscape. SAHARA: Journal of Social Aspects of HIV/AIDS Research Alliance. 2015;12(1):18-29.

6. Martínez-García M, Hernández-Lemus E. Health systems as complex systems. 2013.

7. Pincus D, Metten A. Nonlinear dynamics in biopsychosocial resilience. Nonlinear dynamics, psychology, and life sciences. 2010;14(4):353.

8. Nowak A, Andersen JV, Borkowski W. Dynamics of Socio-Economic systems: attractors, rationality and meaning. 2015.

9. Jones EC. Generative social science: studies in agent-based computational modeling. Journal of Ecological Anthropology. 2007;11:76.

10. Paley J. The appropriation of complexity theory in health care. Journal of Health Services Research & Policy. 2010;15(1):59-61.

11. Burman CJ, Aphane MA. Complex adaptive HIV/AIDS risk reduction: plausible implications from findings in Limpopo Province, South Africa. SAMJ: South African Medical Journal. 2016;106(6):571-4.

12. Burman CJ. The Taming Wicked Problems Framework: A plausible biosocial contribution to ‘ending AIDS by 2030’. TD: The Journal for Transdisciplinary Research in Southern Africa. 2018;14(1):1-12.

13. Burman CJ, Aphane M. Community viral load management: can ‘attractors’ contribute to developing an improved bio-social response to HIV risk-reduction? Nonlinear Dynamics Psychol Life Sci. 2016;20(1):81-116.

14. Gilstrap DL. Strange attractors and human interaction: Leading complex organizations through the use of metaphors. 2005.

15. Burman CJ, Moerschell L, Mamabolo R, Aphane M, Delobelle P. Re-imagining decision making: addressing a discrete social driver of HIV/AIDS through the lens of complexity science. African Journal of AIDS Research. 2015;14(1):75-84.

16. Burman CJ, Aphane M. Complex HIV/AIDS landscapes: reflections on how ‘path creation’influenced an action-oriented intervention. Systemic Practice and Action Research. 2017;30(1):45-66.