Modeling multiple orders of adaptivity from a higher-order adaptive dynamical system perspective

Complex real-world processes often function like complex dynamical systems. Such dynamical systems are inherently adaptive in the sense that not only their variables but also their characteristics can change over time. Furthermore, in many cases the characteristics of a (first-order) adaptation process itself can also change over time, which enables second-order adaptation for context-sensitive control over the first-order adaptation. Moreover, in certain circumstances even more orders of adaptation play a role. In this paper, a generic architecture for different orders of adaptation will be discussed, and illustrated by examples from multiple scientific disciplines. These examples cover first- and second-order adaptivity varying from plasticity and metaplasticity considered in neuroscience and controlled organisational learning in management science to bonding and adaptivity of it considered in social psychology (first- and second-order adaptation). Furthermore, higher orders up to fifth-order adaptation are covered as considered in evolutionary biology, in genetics and in epigenetics and their effect on mental disorders. It is shown how a network-oriented modeling approach based on self-modeling networks can be used to obtain a neat and transparent declarative description for multiple orders of adaptation within one overall temporal-causal network model according to different levels of self-modeling. Moreover, it is demonstrated that any smooth (higher-order) adaptive dynamical system can be modeled according to this architecture.