Autopoiesis and the System Boundary
by Nathaniel Virgo.
I would like to ask this community a question which I think is quite central to a lot of discussions that are taking place at the moment. It has to do with the role of an organism’s physical boundary in the autopoietic theory. I think a lot of tension arises from differing assumptions about this issue and I would like to draw people’s attention to some key parts of the primary literature which support a view that, I think, differs from many more recent interpretations.
I hope that this can contribute to the current ongoing discussion about the relationship between Developmental Systems Theory and autopoiesis. It’s also relevant to some discussions that took place at ALife XI this week, including Mike Wheeler’s talk on whether autopoiesis and the idea of extended cognition are compatible, and the debate about whether the model I presented can be a model of autopoiesis if the spots’ boundaries are blurred rather than distinct.
In Autopoiesis: The Organization of The Living (1972), Maturana and Varela introduce us to the concept of homeostatic machines. These are defined as “machines which maintain constant, or within a limited range of values, some of their variables,” a definition which will be familiar to most of us. However this definition is followed by an important clarification which seems to me to be fundamental to how the rest of the theory is to be interpreted. Since the clarification of this definition is so important I will quote the whole paragraph:
There are machines which maintain constant, or within a limited range of values, some of their variables. The way this is expressed in the organization of these machines must be such as to define the process as occurring completely within the boundaries of the machine which the very same organization specifies. Such machines are homeostatic machines and all feedback is internal to them. If one says that there is a machine M, in which there is a feedback loop through the environment so that the effects of its output affect its input, one is in fact talking about a larger machine M’ which includes the environment and the feedback loop in its defining organization. (section 1.2.a)
Thus if we wish to see a thermostat as a homeostatic machine that regulates the temperature of a room, it is not correct (in to the autopoietic language) to think of the thermostat as being the box on the wall that is connected to a heater and contains a thermocouple, because this machine (machine M) has a feedback loop that runs through the environment. When the temperature drops the thermocouple breaks a connection, which causes the heater (which is in the environment) to be switched off, causing the temperature to drop again. Since the thermostat relies on this feedback loop for its operation, we should actually define the thermostat as a larger machine (machine M’) which includes the heater, the air in the room, and the feedback loop that passes through them.
Why is this so important? The above quoted paragraph is positioned directly before the definition of an autopoietic machine is spelled out [see footnote 1], and just below that we are given the following key statement:
Therefore, an autopoietic machine is an homeostatic (or rather relationstatic) system which has its own organization (defining network of relations) as the fundamental variable which it maintains constant. [see footnote 2]
Autopoietic systems, then, are to be seen as homeostatic machines. It follows that their definition must be expanded in the same way if they rely on a feedback loop that runs through their environment. The implications of this cannot be overstated.
Let us take the excellent example that Mike Wheeler gave in his talk of an earthworm. Earthworms build tunnels, using sticky secretions to hold them open. I can’t find a reference for this right now but I think these secretions help to digest the soil, so that they move back and forth along the tunnels digesting their food again and again. In any case, earthworms have behaviours that modify their environment, and they rely on the effects of these behaviours to maintain their organisation.
So let us try to see an earthworm as a homeostatic machine whose fundamental variable is its own organisation. First let us take the boundary of this machine as being the physical boundary defined by the worm’s skin. There are many feedback loops within this boundary that contribute to this homeostasis. But the worm also builds tunnels using its secretions and then later uses those tunnels to travel along and find food. This is a feedback loop which runs through the environment, and we have been told what to do in this situation: our previous definition of the worm (system W, defined as being bounded by the worm’s skin) must be expanded to a larger system (system W’) which includes the tunnel, the secretions and the feedback loop that runs through them.
The following conclusion seems inescapable: in Maturana and Varela’s framework the network of processes that define the worm as an autopoietic system is not coextensive with the boundary of the worm-as-a-physical-object, it is much bigger. This will be the case with most if not all organisms. As soon as an organism’s homeostasis of organisation relies on a sensorimotor loop there will be feedback loops that run through the environment, and those parts of the environment will have to be included in the autopoietic system. Although the worm does have a physical boundary, and this is important in its definition, it is only the boundary of the worm-as-a-physical-object. It not the boundary of the autopoietic-system-that-defines-the-worm, because the boundaries of that system are much wider.
When I first started reading about the autopoietic theory there were many statements that seemed very strange to me, but with this interpretation they seem much less so: it is not strange to say that changes in an autopoietic system can only be caused by things internal to that system, because the system is defined to include all the relevant causal relations, whether they be within the physical boundaries or not. It is not strange to say that autopoietic systems have no inputs or outputs for the same reason: what we usually think of as inputs to a sense organ or outputs from a muscle are actually internal to the system, because the system is defined to include a large proportion of what we usually think of as the environment.
It’s worth noting that with this interpretation the autopoietic theory starts to look a lot more like the approach of Developmental Systems Theory: it’s about including all the relevant parts of the environment in the causal story that we tell about the organism.
There is one slight problem with this approach, which came up in the discussion at EPIROB about Developmental Systems Theory, and that is the problem of when to stop. It seems like there might always be a system external to the system through which an important feedback loop might run. Should the autopoietic system that defines a single worm include all the plants on Earth, which re-cycle its respired carbon dioxide back into oxygen, a process which it ultimately relies on to maintain its organisation? It seems to me that it might sometimes be useful to say that, and other times not. I believe that this problem is not insurmountable.
However, it seems to me that in many modern interpretations of autopoiesis, and perhaps even in later texts by Maturana and Varela, that this definition of an autopoietic system as extending beyond the physical boundaries of an organism is not present. Later versions of the official definition introduce a physical membrane, and one gets the impression that this membrane is supposed to define the boundary not just of the unity that the system constitutes, but of the system itself. The assumption seems to be that the autopoietic system that defines an organism is a network of processes bounded by the physical boundary of the organism that maintains its own organisation.
The trouble is that I find it hard to understand a lot of the autopoietic theory when interpreted this way. If we try to re-construct the story of the worm it seems quite problematic. It is now defined by a network of processes whose boundary is the physical boundary of the worm-as-a-physical-object. It can perturb its environment by leaving sticky secretions, and it can later come back and be perturbed by the presence of a tunnel, but it’s hard to find the language to express the causal relationship between these two things, because we’re told that there cannot be causal relations between things inside the system and things outside it.
So the question is, what do you think? I’m interested to know whether current researchers read Maturana and Varela the same way I do, as saying that autopoietic systems extend beyond the physical boundaries of the organisms they define. I think that resolving this question of the role of the organism’s physical boundary is important to a lot of current debates, and it would be really good to hear a few opinions on the issue.
[footnote 1] the full definition given in this text is as follows. Note that this version of the definition hinges on the constitution of a concrete unity in space but does not specify that this unity must be bounded by a distinct membrane.
An autopoietic machine is a machine organized (defined as a unity) as a network of processes of production (transformation and destruction) of components that produces the components which: (i) through their interactions and transformations continuously regenerate and realize the network of processes (relations) that produced them; and (ii) constitute it (the machine) as a concrete unity in the space in which they (the components) exist by specifying the topological domain of its realization as such a network.
[footnote 2] as I understand it, Ezequiel’s position is that this does not directly follow from the definition but requires a concept of adaptivity to be introduced as well; in that case I will assume that we’re talking about autopoiesis plus adaptivity rather than bare autopoiesis, since my point is about the interpretation of autopoietic systems as homeostatic machines.
For what it’s worth my personal feeling is that this quoted statement puts across the underlying idea of autopoiesis far more clearly and concisely than any of the attempted formal definitions that Maturana and Varela have given us; and that one does not lose much, if anything, by taking homeostasis-of-organisation as the defining feature of autopoiesis rather than a consequence of it.