Walter Fontana and Leo W. Buss
Self-maintaining natural systems include the global climate system, all living organisms, many cognitive processes, and a diversity of human social institutions. The capacity to construct artificial systems that are self-maintaining would be highly desirable. Yet, curiously, there exists no readily identifiable scientific tradition that seeks to understand what classes of such systems are possible or to discover conditions necessary to achieve them. Given the ubiquity of such systems naturally and the desirability of self-maintenance as a feature of design, any credible approach to establishing such a tradition merits serious attention.
We have recently developed and implemented a framework for approaching the problem (Fontana and Buss 1994a, Fontana and Buss 1994b). It is based on the premise that the constituent entities of a self-maintaining system characteristically engage in interactions whose direct outcome is the construction of other entities in the same class. Self-maintenance, then, is the consequence of a constructive feed-back loop: it occurs when the construction processes induced by the entities of a system permit the continuous regeneration of these same entities (Varela 1974). The specific functional relationships between entities which collectively insure their continuous regeneration, we define as an organization. A theory of organization, so defined, is a theory of self-maintaining systems. A prototypical instance of entities are molecules. And organisms are a particularly interesting class of self-maintaining systems generated by their constructive interactions. The atmosphere is another example. And so, perhaps, is the sun at the nuclear level.
The overarching long-term goal of our program is to develop a formal understanding of self-maintaining organizations. Our efforts in doing so, which we summarize here, have led us to appreciate a fundamental problem in methodology: the traditional theory of "dynamical systems" is not equipped for dealing with constructive processes. Indeed, the very notion of "construction" requires a description that involves the structure of objects. Yet, it was precisely the elimination of objects from the formalism that make dynamical systems approaches so tremendously successful. We seek to solve this impasse by connecting dynamical systems with fundamental research in computer science, whose theoretical foundations are about "objects" and their constructive interrelations. Our long-term goal, then, becomes equivalent to the task of expanding dynamical systems theory to include object construction, to become what we have come to call constructive dynamical systems (Fontana and Buss 1994a).