ABSTRACT
The dynamical behavior of complex information-processing systems, and how this behavior may be improved by natural selection, or by other learning or optimizing processes, are issues of fundamental importance in biology, psychology, economics, and, not implausibly, in international relations and cultural history. In coevolving systems, fitness landscapes themselves deform due to coupling between coevolving partners. Conditions for optimal coevolution may include tuning of landscape structure for the emergence of frozen components among the coadapting entities in the system. Organisms are the paradigmatic examples of complex systems which patently have evolved, and hence now do fulfill the requirements of evolvability. Real coevolution confronts not only adaptive moves by coevolving partners, but exogenous changes in the external "world" impinging upon each partner. The structure and couplings among landscapes reflect the kinds of entities which are evolving and their couplings. Natural selection or learning may tune both such structures and couplings to achieve systems which are evolvable.
