The Fitness Landscape and Co-evolution
"Fitness" in this sense is the suitability of the strategy that a given agent pursues to achieve its purpose: its own survival or success compared to the relative suitability of the strategies used by other agents in the system of their fitness. A three-dimensional plot of the fitness of all the potential strategies that could be employed by the agents that compose the system environment gives us the fitness landscape. This landscape consists of a topography of peaks and valleys whose elevations correspond to the advantage or disadvantage that a given strategy offers compared to the other potential strategies. The peaks represent strategies that lead to success; the valleys represent strategies that lead to extinction. (See Waldrop 1991, Lewin 1992, Casti 1994, Goodwin 1994, in The Library. See also the concept of "The Tapestry Theory of Knowledge" in Loverde, The Hidden Soul of Capitalism Through Dynamic Markets Leadership, 2013.)
As agents in the system respond to each other’s behavior in order to improve their ‘fitness’, or comparative advantage within the system, their actions influence the shape of the landscape. At the same time that individual agents in the system modify their own behavior, other agents make behavior modifications of their own in response (co-evolution). Some strategies might involve incremental movement toward an existing peak, others may require traversing a 'valley' in order to find a path to a given peak. As agents engage their own strategy and realize how others may be duplicating or attempting to offset their own efforts to improve their situation, they often create alliances with others within the system who are attempting to do the same, essentially self-organizing into emergent patterns (with no blueprint or external guidance) of behavior that tend to maximize the likelihood of mutual success, and thus the fitness of the system they comprise (Kauffman, 1993 ref). As an agent journeys to the peak in mind, the actions of other agents in the meantime may cause what had been an advantageous position to become comparatively less so. Alternatively, from the new 'peak' position, an agent may become aware of other existing peaks that were not visible from its previous location on the landscape. Thus the co-evolution of the various agents generates continuous shift throughout the system landscape.
In addition to constantly shifting standards of strategies and advantages, agents have to consider the ethical rightness. In CAS the lack of simple blueprints or predictabilities are what expose the weakness of one ethical model, Utilitarianism, that depends on predicting consequences in order to consider what is right. Therefore, agents who are operating either for their own benefit or for the benefit of a whole system that might yet emerge and self-organize are constantly making ethical decisions, regardless of the limitations of Utilitarianism. The main alternative is the ethical model of Deontology that does not require predictions to assess rightness; instead, Deontology makes ethical judgments based on higher-order principles with claims to universalizability; for instance, if slavery is wrong based on an ethical principle of the right to freedom, liberty, or dignity then it doesn't necessitate the probability of a Civil War to determine a decision that slavery is wrong. Another important ethical model is that of virtue or ethical character, where integrity is crucial. Therefore, when a Board of Directors needs to provide guidance for Complex Adaptive Systems in its company, Deontology can be used to assert what is universalizable; and Integrity Theory can be used to promote agents who have wisdom in their judgment.
In other words, such systems operate in a manner that constitutes learning and adaptation, and, in fact, not only allow but tend to require continuous learning and adaptation in order to stay competitive.