Modeling Autopoiesis and Cognition with Reaction Networks

Maturana and Varela defined an autopoietic system as a self-regenerating network of processes. We reinterpret and elaborate this conception starting from a process ontology and its formalization in terms of reaction networks and chemical organization theory. An autopoietic organization can be modelled as a network of “molecules” (components) undergoing reactions, which is (operationally) closed and self-maintaining. Such organizations, being attractors of a dynamic system, tend to self-organize—thus providing a model for the origin of life. However, in order to survive in a variable environment, they must also be resilient, i.e. able to compensate perturbations. According to the “good regulator theorem” this requires some form of cognition, i.e. knowing which action to perform for which perturbation. Such cognition becomes more effective as it learns to anticipate perturbations by discovering invariant patterns in its interactions with the environment. Nevertheless, the resulting predictive model remains a subjective construction. Such implicit model cannot be interpreted as an objective representation of external reality, because the autopoietic system does not have direct access to that reality, and there is in general no isomorphism between internal and external processes.