Harnessing Complexity by Axelrod and Cohen is an excellent introduction to Complex Adaptive Systems, specifically focusing on what this means for management. There is a lot in this book, and I can’t even scratch the surface here, but it is worth sharing a few points around three key interwoven processes: variation, interaction, and selection.
Variation is the diversity in the system. This broad definition of diversity includes the properties, capabilities, and strategies different agents use within the system. It is best to think of it from a behavioral standpoint and focus on the strategies knowing that the preferred strategies will be connected to the unique properties and histories of the agents in the system.
Variation provides the raw material for adaptation…when the world is changing or the current agents are far from the best possible, variety can have value, and homogeneity may be a hindrance.
The balance between variety and uniformity is “an important trade-off principle, usually called exploration vs exploitation.” Too much variety in the system is referred to as Eternal Boiling.
“In such a state, any potentially valuable structures are broken apart before they can effectively put to use.”
At the other extreme is premature convergence.
“Premature convergence occurs when needed variability is lost too quickly.”
With variation, there is a balance between lots of variation, which is great for exploring new possibilities, and minimal variation, which is great for efficiency.
As a result, as a manager, two actions you can take are:
Arrange organizational routines to generate a good balance between exploration and exploitation.
Link processes that generate extreme variation to processes that select with few mistakes in the attribution of credit.
When dealing with a complex adaptive system, a key question is “what should interact with what, and when?”
How things agents and strategies interact in a complex system determines a lot of how that system behaves. The two facets of interaction are proximity: “how agents come to be likely to interact with each other” and activation: the factors that impact the timing of an agent’s activity.
The principle mechanisms available to change interaction patterns from outside include the construction and operation of barriers to interaction—or removal of such barriers.
These barriers can be real or conceptual. How you think of your team and who is invited to “team” meetings is just one simple form of boundary that is easy to play with.
As agents interact, they will copy behaviors from each other.
The mechanism of copying the interaction patterns of other agents passes along vital social knowledge and allows an agent to adapt, without requiring an explicit understanding of very complex social systems.
However, there is a downside.
problems can arise when interaction patterns are transferred to new contexts, since the selectivity of a more precise theory is not available to sort out which features should be modified and which retained.
Not every strategy will work in every situation. The risk is that transferring strategies from one context to another could lead to poor strategies.
It may seem a paradox that as individual agents experience more diverse contacts, the system can become less diverse.
Early innovations spread too fast, and variety that can provide later improvements is lost—the phenomenon we have called premature convergence.
There are four key actions that you can take:
Build networks of reciprocal interaction that foster trust and cooperation.
It can be really useful to think about how your team members interact with each other and how they interact with others in and outside of the organization.
Assess strategies in light of how their consequences can spread.
Strategies spread through interaction, and different strategies may be easier or hard to spread throughout a network. It is important to think about strategies not just based on how effective they are but also on how well they can spread.
Promote effective neighborhoods.
There is value in having the right degree of coherence in a group, and this will depend on context. It can help find the right degree of similarity to foster in those coordinating.
Do not sow large failures when reaping small efficiencies.
Increasing connections (tearing down silos) can be beneficial to create efficiencies and transfers of strategies, but there is a downside. To use an analogy, a virus is more likely to spread in a tight network than in a sparse one. It is important to find the balance between interaction and separation.
Selection is the process by which a complex system evolves by choosing the frequency of various agents and strategies.
To know what to select, it is important to define success,
First, it is valuable to appreciate that performance measures are defined within the system.
These definitions of success are not always explicit. Many cues for what is successful and what isn’t are subtle. It is important to note these rewards (raises, promotion, praise, better assignments) all happen within the system.
Second, how success is defined affects the chances for effective learning.
When success is measurable only rarely, new measures with a faster tempo can speed learning, even if they do not perfectly reflect the longer-term goal.
Whenever outcomes are better or worse than expected, the experience can help to revise evaluation criteria so that, in the future, the attribution of credit will produce better outcomes
It is also important to note that causality is hard to know in a complex system. Behaviors correlated with success may get copied despite lacking a causal relationship.
Often what will lead to success will be strategies, but those strategies are used by agents. Often the agent is rewarded and thus both the successful strategies and unrelated strategies are more likely to be copied. If you can reward just the specific strategy it is often more transferable, but there is the risk of loss of context.
Agent selection often works on longer time scales—faster is not always better—and preserves variation and context. Strategy selection isolates key patterns that can be more easily and rapidly copied.
There are three major risks to determining what caused the success within a complex system:
- the mistake of crediting or blaming a part when a larger ensemble is responsible,
- the mistake of attributing credit or blame to a particular ensemble of factors when in fact a different ensemble is responsible, and
- the mistake of crediting a misconstrued strategy, where the action involved produced success, but the conditions in which the action should be taken have been misunderstood.
As a leader, there is a much higher risk that what you do will be copied:
First, a leader can sometimes set standards that provide incentives for others to copy. Second, a leader’s actions or performance measures are typically seen to be successful and hence worth emulating. Third, a leader may set an example that helps establish beneficial norms in a community.
To support appropriate selection there are two key actions you can take:
Use social activity to support the growth and spread of valued criteria.
Look for shorter-term, finer-grained measures of success that can usefully stand in for longer-run, broader goals.