A part of complexity science that deals directly with nature-inspired evolutionary processes involving interaction in the fitness function. A definition I've used in the past is that these algorithms involve an evolutionary dynamics in which one or more types of agent playing two or more distinct roles interact with measureable outcomes that impact the agents' future evolution. interactive domainsinteractive domains
A collection of one or more functions, called metrics, of the form $p\colon X_1\times X_2\times\cdots\times X_n\rightarrow R$, where
each $i$ with $1\leq i\leq n$ is a domain role
an element $... are one possible precise formulation of the underlying data needed; these encompass the payoff matrices used in game theory, but are simultaneously less constrained and more general. Because there is an evolutionary dynamics at work, the algorithm has a state consisting either of one or more populations of agent or one or more distributions over agents.
Through my PhD dissertation work I came to understand that the mechanism by which the impact of interaction outcomes on future evolutionary trajectories is realized via the interactions' function as measurements. How any two agents compare can change fundamentally (e.g., change order) depending on the interactions in which they participate. I developed a way of capturing that information in what I termed coordinate systems, a notion that led to DECA as a published algorithm. Coordinate systems will be multi-dimensional in general, an observation that can be used to explain a number of pathological dynamics such as cycling, overspecialization, or lack of robustness that have been observed in coevolutionary algorithms and limit wider applicability.
