A living organism may be characterised as a self-sustaining network of processes. This is a common theme running through many distinct approaches, including those of Maturana and Varela, Rosen, Gantí, Fontana and Buss, Kaufmann, and Beer, to name but a few. I argue that understanding the central notion of a process is crucial in order to properly formalise this idea. I argue that processes are best understood at a macroscopic, statistical level of description, and that this will introduce an unavoidable degree of observer-dependence into any formalism. This observer-dependence is akin to the observer-dependence in classical thermodynamics, and is not necessarily problematic.
I illustrate these ideas with some simple simulation-based examples in which self-sustaining and self-individuating networks emerge in abiotic systems. From these examples I argue that (i) there is no true distinction between the “behavioural level” and the “metabolic level”, with the self-sustaining networks of even the simplest agents crossing between the two; (ii) the processes in the network should not be thought of as being physically “inside” the agent; and (iii) in some cases it may not be possible to define processes separately from the individuals they are part of.
I conclude that we should be thinking not about whether something is an agent in this particular sense, but about whether it is productive to view it as an agent. This is analogous to Dennet’s intentional stance, but it deals with the physical, self-producing elements of living agency as well as the behavioural level.
Agency in the Physical Sciences 