Autopoiesis and Coevolution

by Chris Lucas

"The stability of the internal medium is a primary condition for the freedom and independence of certain living bodies in relation to the environment surrounding them."

Claude Bernard, Leçons sur les Phénomènes de la Vie Communs, 1878-1879

"If we are to understand a newer and still evolving world; if we are to educate people to live in that world; if we are to legislate for that world; if we are to abandon categories and institutions that belong to a vanished world, as it is well-nigh desparate that we should; then knowledge must be rewritten. Autopoiesis belongs to the new library."

Stafford Beer, Preface to Autopoiesis, 1980

Introduction

Autopoiesis is based on the way living systems address and engage with the domains in which they operate. This biologically based theory, introduced here, (originated by Maturana and Varela) defines life as the ability to self-produce, rather than as (conventionally) the ability to reproduce. Like complexity theory it is a systems perspective, and is applicable to brains and societies as well as to biology and artificial life. In its original form it was applied to cognition, and replaces an external objective view of this subject with an internal relativistic understanding, in terms of an embedded observer.

Structural Determinism

Only two type of structural change are possible in living organisms, changes of state that preserve identity, or disintegration (death). External perturbations trigger the changes to the organism but do not themselves determine them. The available states of the organism determine which environmental triggers can be recognised and which will disintegrate it, but the available states occur by a process of self-organization. This is the concept of 'structural closure or determinism' that forms a main theme in autopoietic thinking. Note that this means that we do not 'map' our environment but just respond to a subset of it - a simplification required by the cybernetic Law of Requisite Variety.

Instructive interactions (traditional imposed learning) are said by Maturana to be incapable of being subjected to scientific procedures. Like molding clay, they leave no trace of the original structure. Thus the standard neo-Darwinist idea that the environment imposes order on organisms is at best metaphorical. For example, the light entering our eye does not 'cause' the photochemical release that occurs, that mechanism must already exist, light just triggers it. This ties in with the complexity view that selection acts on systems whose structure has already self-organized.

Structural Coupling

An organism undergoes a history of perturbations from its environment (also seen as a dynamical system at a higher level, subsuming the organism) which trigger its own state trajectories. If these triggers are regular, then the organism also has regular state cycles, if continuously changing then novel trajectories will occur. If the organism affects (triggers) the environment in turn then we have the coevolution of two structurally coupled systems.

If these triggers result in state changes that involve component changes (rather than just interactional ones) then we have adaptation. This plasticity in structural topology at both state and component levels is what is called 'structural coupling' and can operate in both directions. By this process the medium or environmental organization becomes coordinated with that of the organism. But note that there is no necessary informational or semantic commonality between the two, each reacts to the other on its own terms.

Self-Production

The components of the organism are regarded as a form of autocatalytic set. In other words the components take in food (lower level components) from the environment and by a dissipative process (using energy) act on each other in such a way as to recreate themselves dynamically (unlike machines with fixed parts). This closure of the system allows it be bounded, to isolate itself from the outside world and become a self-sustaining constant (homeostatic) system. This definition of life is far better that the systemic illogicality of defining it as the reproduction of a passive gene. A living system is a ongoing process that self-defines and self-maintains its form, reproduction is not a necessary function of this.

Autopoietic closure is a systems property and is thus not restricted to biological systems. It gives Autonomy, Phenomenal Distinctions - an emergent composite level as well as a part level (these are not logically deducible from each other), Adaptability - tracking its environment in the present (ontological) and Adaptation - divergent trajectories of parallel systems (some disintegrate). Thus reproduction and natural selection is a domain orthogonal to self-producing 'being' or existence itself.

Learning and Instinct

The organism's current state is due to its history of structural coupling. This base or instinctive state is a combination of evolutionary and developmental processes. Learning is also a form of structural coupling, so is indistinguishable from instinct. Distinctions between evolutionary, developmental and experiential structural change are thus those of timescale and not of type, and this suggests that they can be modelled by a common mechanism.

The independence of system states from environment states allows for decoupling, seen (from outside) as inappropriate behaviour - often called 'instinct' or 'mistake'. This relates to systems that are internally dynamically self-consistent (able to follow their own processes) yet do not act on their environment in a way consistent with some external understanding. This insight suggests that many alternative and equally valid 'human systems' can exist without being regarded as 'faulty' - an idea that has many implications for medical interventionism aimed at 'normalisation'.

Consensus Domain

This is the structural coupling between two autopoietic and structurally plastic systems. The interactions between the two systems are closed on triggers but are open on the systems that realise them (organisation is independent of the parts that create it). This means that we can substitute functionally equivalent systems without affecting the coevolutionary result. Thus artificial lifeforms would be indistinguishable from natural ones from a behavioural point of view.

Further, because any individual is organisationally closed and cannot distinguish the source of its triggers, then the 'reality' that we create bears no relationship to the external world. We do not model or describe a reality, simply a consensus. Interacting systems serve as a source of (cybernetically) compensatable perturbations to each other, but do not share explicitly any external common reality.

Cognition and Language

Our discrimination of our environment (in autopoietic thought) is constrained by structural limitations, the participant's possible organizational modes. Each mode forms a subset of the structurally determined set of available actions, and this set changes over time with the system's structural coupling to the environment (learning, triggers). Each cognitive entity therefore occupies a niche or specific 'cognitive domain' closely linked to its own history of structural coupling. In autopoiesis, calling the brain an 'information processing device' is patently wrong - that is a description from an external observer viewpoint and not one experienced by the organism. Information processing is an external analogy (like attributing emotions to robots) and does not relate to an internal description of a real symbolic manipulation system.

Language is not therefore an 'instructive interaction', passively moving information between two entities, but a mutual orientation of the participants to a common consensus domain - a linking of their separate cognitive domains. We can only agree as far as our structural options let us, what an observer assumes (from their separate domain) to be a universal agreement about reality may be no more than a common history of structural coupling - a correlation or coordination between two discrete entities achieved over a period of mutual plastic changes. This viewpoint generalises language to include many other forms of signalling, and thus makes all semantics contextual and not absolute.

Identifying Autopoiesis

Six criteria are given to determine if a system is autopoietic:

1.   Does the system have identifiable boundaries ?

2.   Does it have constituent elements or components ?

3.   Is it mechanistic (subject to cause and effect) ?

4.   Are the boundaries self-produced ?

5.   Are the components of the boundaries self-produced ?

6.   Are the rest of the components self-produced ?

Under these criteria, not only are biological organisms autopoietic, but so is cognition, society and many institutions within it.

Forms of Coevolution

Coevolution relates to the two-way interplay between the organism and aspects of its environment and can occur at various levels. Firstly the organism can affect the physical environment, by changing the adaptive pressures on itself (e.g. by moving around, digging holes), thus the physical environment should not be regarded as a static object. Secondly physical changes in this environment can affect the organism (e.g. weather changes) leading to adaptation. Most forms of coevolution however will occur with respect to other organisms, and again there are several aspects to this. The interactions can be with members of its own species, either competitive or cooperative, with 'prey' species (lower on the food chain), with 'predator' species (higher on the food chain), with neutral species (at any level) and with kin.

All these interactions may be multi-faceted, i.e. they can affect several interacting values or needs in each organism. A particular set of relevant interactions, rarely considered, relates to synergic coevolution, where a number of organisms enter in a mutual arrangement such that the net benefit to all exceeds their individual costs. The also takes many forms including symbiotic, social and ecological networks. In general all these effects will be present to some degree in any complex system, including economic and cultural ones. This necessitates a more holistic approach if we are to understand which mutual structural changes are taking place, which niches are being dynamically created, and their relative stabilities.

Conclusion

Autopoiesis, with its stress on action within an environment helps us to understand life at all levels. This relates to the situated or selected self-organization mode of complexity thought, which considers the coevolution of system and environment. Whilst autopoiesis usually does not incorporate the complexity concept of dynamical attractors it uses the same idea of limited flexibility due to structural connectivity, along with the need to change structure if we are to develop new modes of behaviour. No mechanism is suggested however to drive the structural changes and in this respect complexity thought goes beyond this field, allowing for internal mutation or recombination to generate emergent options for subsequent testing against environmental response.