Mechanistic explanations and structure-determined systems: Maturana and the human sciences
Burghgraeve P. (1992) Mechanistic explanations and structure-determined systems: Maturana and the human sciences. In: Van de Vijver G. (ed.) New Perspectives on cybernetics: Self-organization, autonomy and connectionism. Kluwer, Dordrecht: 207–217. Available at http://cepa.info/2742
Table of Contents
1. The origin of characteristics
2. Making distinctions
3. Making distinctions leading to mechanistic explanations
4. Making differences between entities from a mechanistic point of view
5. The identification of structure-determined systems
6. Explanation in human sciences
The notion of mechanistic explanation is given a very specific meaning by Maturana. Does it influence research in human sciences? In this article an attempt is made to answer this question.
When first confronted with Maturana’s description of a living system, his general point of view made sense to me.[Note 1] I did not especially try to know everything he said or wrote in detail. This is probably the reason why it took me some years before I decided to read the 1978 article ’Biology of Language: The Epistemology of Reality.’[Note 2] I decided to read it, because it is claimed that this article is the real clue to Maturana’s thinking. After reading it, my conclusion is that it is a very difficult article. I didn’t come to formulate a personal view on it.
Two or three times I started to read it again. The more I looked at it, the more I became convinced that, if Maturana proves right in his approach of systems in general and living systems in particular, our outlook on human sciences has to change drastically. This impression is shared by other researchers and is generally expressed by stressing the fact that no point of view, theory, etc., on man and society can objectively be preferred to any other point of view, theory, etc.
So, what I am writing down now, is an attempt to follow Maturana’s thinking with the intention to answer the question: does Maturana’s point of view make a difference to further research within the field of human sciences?
First I thought I would find the necessary arguments when going deeper into the ideas he formulates concerning living systems. To my own surprise I didn’t need to go into particulars with regard to autopoietic systems. A sufficient reason already comes to the fore when looking at his description of scientific research.
1. The origin of characteristics
Being involved as I was with AI, I decided to start looking at the paper from the angle of functionalism. Functionalism itself is not an easy concept. It has several slightly different meanings, which are too complex to take into account within the context of this article. I use this approach only as a starting point and a reference. Hence, I start from a very general definition of functionalism as a specific way to study systems.[Note 3]
A functional approach to study systems implies the analysis of the system into parts, the description of the specific relations between the components and the description of the interaction or processes taking place between them, with the intention to explain why the system behaves as it does. Usually research based on a functional approach includes the specification of the relations between certain parts of the system and the environment of the system.
If we take this general definition as a starting point, we can use Maturana’s approach to scientific explanations and characterize a functional system description as an attempt to give a mechanistic explanation of a particular system.
At first I thought this characterization of a functional description to be quite correct, since Maturana states that it is only in mechanistic explanations that the observed system is treated as a composite unity. But I became less sure of this when trying to proceed with the interpretation, using my definition of functionalism as a leading idea.
Maturana states that he distinguishes between two standard scientific explanations. On the one hand he introduces the vitalistic explanation, which explains the characteristics of a system by stating that they are found among the properties of a component of the system. On the other hand, the mechanistic explanation explains characteristics and properties of the system as the result of the specific interaction of the components of a system and not as characteristics of components as such.
Classifying the research as mechanistic, means that it is necessary and sufficient to state that the system under study only behaves as it does because all the components contribute to it. Therefore, a mechanistic explanation necessarily needs a decomposition of the system into components and their interrelations. A vitalistic explanation doesn’t need this decomposition, because the specific interaction between the specific components doesn’t play any part in the description of the characteristics of the system.
Can we state that a functional approach which tries to explain why a system behaves in a particular way, always leads to a mechanistic explanation? Following Maturana, I am inclined to give a negative answer to this question. We could certainly imagine a model of the brain, starting from the assumption that the property of humans to remember is the sole result of the property of a memory-unity, that the property of thinking is the sole result of a thinking engine such as a computation-unity, etc. And this way of looking at human cognitive abilities is not as far-fetched as one would think, because this has been a fairly common way of speaking metaphorically about human cognitive abilities since the beginning of computer age.
The example is rather an obvious one and I use it to make it clear that one could easily think that one is explaining something mechanically, but that this doesn’t imply that this seemingly mechanical way to explain a phenomenon is what Maturana means when he is talking about a mechanistic explanation. The obvious conclusion is that the decomposition in parts is not a guarantee for a mechanistic explanation, but that it is vital that no characteristic of the complete unity should be explained without reference to the interrelations of the parts. Hence, not all functional analyses resolve into a mechanistic explanation.
Assuming that I understand the difference between vitalistic and mechanistic explanations correctly, I could make a risky generalization and state that from Maturana’s point of view, every methodological stand, up to this moment, proposes obscure views as to the choice of the origin of characteristics. In other words: researchers don’t seem to know exactly why and when an explanation is a mechanistic one and why and when an explanation is a vitalistic one.
The obvious question now is why they didn’t have a clear view of the difference between those two forms of explanation.
2. Making distinctions
A unity comes into being through operations of distinction. Those operations assign properties to something. Making a distinction, and as a consequence assigning properties and hence seeing something as a unity, depends on the observer. Being able to see a ’thing’ is the result of the possibility of the observer to be aware of a difference between ’it’ and another ’thing’.
Two crucial notions emerge here: ’making distinctions’ and ’characteristics’. They seem crucial to me because they play a part in the distinction between mechanistic and vitalistic explanations. So let us examine them more closely.
When I start to analyze something functionally, I want to find the reasons why, or in other words, explain why a system behaves as it does. This means that when I make the distinction between my daughter and my cat, I classify them within two different classes, not because I first dissected them, but because I made a distinction as to their behaviour.
But I could easily imagine other ways to make a distinction between my daughter and my cat. I look at them and without waiting for them to act I see that they definitely haven’t the same physiognomy. Because I am a human being with a particular education, I can find all sorts of other ways to distinguish between cats and human beings. Even when only taking into account the distinctions based on behaviour and those based on physiognomy as a starting point, I could easily realize that not every possible distinction is suitable for a mechanistic approach.
3. Making distinctions leading to mechanistic explanations
Maturana must have been aware of this. Hence we have to look for further specifications of the mechanistic explanation. He offers some help by the way he entitles the second part of his article: ’Operational Characteristics of a Mechanic explanation’.
Organization, structure and space are the three main notions which can shed a light upon this problem. I start with the notion of space, because it seems to me that this notion plays a crucial part in understanding the particular choice of the operations of distinction to make a mechanistic explanation possible.
“Space”, he says, “is the domain of all the possible interactions of a collection of unities (simple, or composite that interact as unities) that the properties of these unities establish by specifying its dimensions.” And further: “Once a unity is defined, a space is specified.” (Maturana, 1978, p. 33).
The moment I decide to point at something as a unity, I can specify its space. It becomes immediately clear that every identification gives rise to two particular spaces, because every unity can be viewed as simple or composite.
Let us try to see what this statement amounts to by alternately looking at a human being as a simple and as a composite unity. As a simple unity, the space of a human being definitely lies at the outer side of his skin. I characterize him, for example, by taking his social behaviour into account, by noticing that he has neither feathers, nor fur etc. As a composite unity, the space of a human being definitely lies at the inner side of his skin. All what is going on inside can be taken as characteristics to establish the human being as a composite unity.
In addition to this central dichotomy, we have to take into account the actual possibilities of an observer to observe. These can change and this change can influence the spaces he is establishing.
Let us now look at the notion of ’organization’. This is one of the main notions playing a part in mechanistic explanations. It is the organization of something that establishes it as a composite unity and organization refers to the specific instrumental participation of components. The composite unity, defined by its internal organization, establishes a subject-dependent mechanistic explanation.
To recapitulate, at the very outset I identify my daughter as a simple unity by establishing an outer skin-space. Depending upon my choice as an observer, I can, for example, use characteristics – being properties resulting from operations of distinctions – that make me look at her as my daughter and no one else, or generally as a human being, or as a mammal, etc., but only if I start to search for inner skin characteristics do I look at her as a composite unity and only then the conditions are fulfilled to start the construction of a mechanistic explanation.
If, up to this point, I understood Maturana correctly, I can say that when choosing, as an observer, a particular set of intellectual tasks as the outer skin characteristics of human beings, it is possible to identify human beings and computers as simple unities belonging to the same class of entities. It seems to me that this is exactly what the Turing test amounts to. But at the same time, this identification, based on the specifically chosen outer skin characteristics, excludes a mechanistic explanation of the identified entities. To identify them from a mechanical point of view, I have to specify inner skin components and relations between components which allow for an organic definition of both entities.
To keep the argument clear, I will give the description of the third main notion later on in the text. First I want to come back to my functional approach. When I say that I want to explain how a unity behaves by splitting it up into parts, I can imagine an objection from someone following Maturana. He or she immediately will attract our attention to the fact that outer skin characteristics and inner skin characteristics are two quite different things. He or she may easily wonder whether it is allowed to say that one explains outer skin characteristics by establishing connections with explanations of inner skin characteristics. The behaviour of the entity as a simple entity emerges from all that is happening inside, but cannot be reduced to a description of a particular inside activity.
4. Making differences between entities from a mechanistic point of view
A mechanistic approach explains what is going on in the inner skin action space of a unity. To decide that something is no longer a member of a particular entity class from a mechanistic point of view, is to state that there is an internal difference between this thing and the members of the entity class.
If I go back to functionalism and artificial intelligence, I can state that a computer-representation functionalism is undoubtedly a mechanistic approach of the thing called computer, but can only be a mechanistic approach of a human brain if they are both – computer and brain – internally identical. Knowing then how Maturana defines organization, can we claim them to be organizationally identical?
I deliberately didn’t go into the notion of ’structure’ in the previous section, because I think that this notion can shed some light upon an additional specification of a mechanistic explanation.
Organization as well as structure are concepts used by Maturana to say something about the way of existing of unities, but it is only the concept of organization that plays a part in the identification of a class of composite unities.
Let us restrict ourselves to the notion of organization, and see whether we can give an account of the brain organization that is equivalent to the one we use to identify a computer. We only have to turn to an arbitrary textbook on computer science and we shall always find the same description of the organization of the computer. The parts of the composite unity ’computer’ amount to: an arithmetic unity, a control unity, a memory or storage unity, an input- output unity and their specific interactions. And without the slightest chance of being corrected, we can state that the mind can be viewed in the same way.
This changes if we take the notion of ’structure’ into account. The structure of a composite unity refers to ’the actual components and the actual relations that these must satisfy in their participation in the constitution of a given composite unity’ (Maturana, 1978, p. 32).
To refer to the actual components and actual relations of a computer, and hence to refer to the structure of a computer before and after installing new software, would mean that we are talking of the same computer with a different structure at different moments in time.
We can use an even better way of presenting this example to stress the difference between organization and structure. The organization of a computer in terms of a Turing machine amounts to the specification of a set of conditionals in the form of a machine table. The structure of this particular computer consists of the actual values we get after, let us say 100 steps, given the aforementioned conditionals. In this formalism, the organizational identification of a computer and the mind could be stated from the moment they have one and the same machine table. To refer to their structure would mean to look at the actual values their conditionals generated after a certain period of time.
In the above ’examples’ I deliberately used ’mind’ instead of ’brain’, because if I start constructing a definition of the organization from a real human brain, it becomes quite difficult to insist on identical organizational definitions of brain and computer.
5. The identification of structure-determined systems
Only because he uses this particular description of mechanistic explanations, can Maturana come to the determination of a class of entities as structure- determined. Structure-determined systems either undergo changes of state ’determining the particular structure at a particular moment’, or they undergo changes leading to a loss of identity.
There would be nothing very spectacular in the specification of this particular class of entities, if the introduction wasn’t accompanied by some penetrating remarks about how changes come about in structure-determined systems.
In the first place, there is the striking remark that changes of state of a structure-determined system can be triggered by perturbing interactions, but that they are not determined by the properties of the entity producing the perturbing interaction. It is very important to understand this perfectly.
Taking again the computer as an example may be quite helpful. Using a function key implies the execution of a particular sequence of instructions. The perturbing action is, in this case, the simple use of a particular function key. When taking the same finger, and using the same, but redefined function key, it becomes quite clear that the particular interaction (using the key), coming from a particular entity (person), with specific properties, doesn’t play a part in the specific developments of the internal structural change. Only the actual structure determines what’s going to happen and the perturbing entity has nothing to do with the actual change. It is only because the computer is such a ’simple’ device, that we can predict the change following the use of a function key.
Applied to a human being this would for instance mean that to say that someone is influenced by someone else, only refers to the fact that the last person is acting as a triggerer of structural change with regard to the first one, but that the triggerer isn’t sure that he can be a triggerer, nor what will be the result of being the triggerer.
If we say of something that it is structure-determined, this necessarily means that we look at the system as a composite entity. We know though, that to look at something as simple or as composite depends on the choice of the observer. Knowing that a system is structure-determined doesn’t prevent the observer to see overall characteristics of the entity. One specific form of characteristic of the entity as a simple entity can be a susceptibility for certain forms of perturbing interaction. This results in particular structural changes only determined by the organization and the previous structure of the system. This can lead to a change in characteristics of the entity, looked upon as a simple entity, etc.
The remark about the absence of influence coming from the properties of the perturbing entities has to be seen in connection with the remark about instructive interaction. Maturana says: “Systems that undergo instructive interactions cannot be analyzed by a scientific procedure… all instructable systems would adopt the same state under the same perturbations and would necessarily be indistinguishable to a standard observer.” (Maturana, 1978, p. 34).
The notion of instruction has again a very specific meaning. Instruction is explicitly coupled to the idea that properties of a perturbing entity could specify the changes of state a system goes through. Those systems are, by definition, not structure determined. The fact that Maturana states that they cannot be analyzed scientifically, because they would be indistinguishable to the standard observer, caused me to have some problems at first. I only could interpret this remark by imagining a system that had escaped from the world of science fiction, that had the means to change itself perfectly and instantly into the perturbing entity. We shall see that the system involved needn’t be so strange and that at least partial instructable systems form an important part of our world.
Living systems are autopoietic. Autopoietic systems are a particular subset of structure-determined systems. They are in fact autonomous entities and their autonomy is a direct consequence of their particular organization which allows the system to arrive at a continuous self-production of the necessary components and relations between components to maintain its organization. Because autopoietic systems form this particular subset of structure-determined systems, they certainly cannot be instructed. Their structural change has nothing to do with the properties of perturbing entities, but has everything to do with their own internal structural situation at a particular moment in their personal history.
Every outer skin property, or, with the words of Maturana, every characteristic of an autopoietic system seen as a simple entity, is the result of an ontologically completely determined change of structure. The autopoietic entity as a simple entity has no means to act upon the completely determined internal structure that is actually responsible for the actual domain of perturbations. It has no means, moreover, to control all the perturbing entities that are made possible by its structure.
6. Explanation in human sciences
We now switch to human beings as autopoietic systems. Instead of following Maturana through the concepts of structural coupling between structurally plastic organisms and without making use of the notion of consensual domain, we can ask ourselves if it is possible to analyze the characteristics of a human being as a simple entity from a mechanistic point of view.
Characteristics of it are for example the use of language, the production of artifacts, etc. This implies two things: on the one hand, physical traces such as buildings, books, pictures, paintings and the very specific outlook of those physical traces, on the other hand, shortlived behaviour that can be the perturbing factor to another structurally plastic organism.
To analyze the physical traces mechanically means that one tries to define the organization of the physical object. The organization of the objects makes it possible to classify them into categories such as books, computers, etc. But the organization doesn’t account for the very particular shape the objects have. This very particular shape is the result of a transfer, of a partial instruction. The properties of the disturbing entity ’human being’ at a particular moment of its existence – having a particular structure, which determines and individualizes its behaviour – are transferred to the entity under consideration. The entity is a frozen behaviour of the human being as a simple entity engraved on a particular physical entity. This is the only way I can imagine instructable systems in our particular world.
Maturana says that instructable entities cannot be analyzed scientifically because a standard observer wouldn’t be able to distinguish those instructable systems, as they and the disturbing entity would be quite alike. But since the instructable entity consists of a frozen instruction, it must be possible to differentiate it. The content of a book doesn’t change any more once it is written down. The entity is not an instructable one any more. And since the physical traces of frozen human behaviour are not submitted to structural change, this content cannot be viewed as a structure-determined entity. The book can be a perturbing agent for a human being, but its frozen behaviour cannot be perturbed by the act of reading.
Hence every example of a physical trace of human behaviour can be differentiated by an observer. Every physical trace can be viewed as a simple entity or as a composite entity. “Everything that is said about the physical trace, is said by an observer to another observer, who can also be himself or herself.” (Maturana, 1978, p. 31). Hence, we can create a “closed cognitive domain in which all statements are, of necessity, subject dependent, valid only in the domain of interactions in which the standard observer exists and operates” (Maturana, 1978, p. 29). Consequently, if we accept that the properties of frozen human behaviour are to be explained by the relations of the components of the frozen behaviour and that they are not to be found among the properties of the components, we start producing a mechanistic explanation of this particular form of entities.
This argumentation cannot be repeated in the case of the short-lived behaviour, although the latter can play an important part in the change of autopoietic systems. Hence, part of the analytical requirements to make the appropriate mechanistic analysis, cannot be fulfilled.
Is there any possibility to use the scientific research of frozen structural states to predict and control autopoietic systems? Yes and no, I should say. Yes, if it would be possible to produce a complete machine table for every second- order plastic structure and achieve complete control over every structure trajectory of every structure-determined system. No, because this isn’t possible for the very good reason that, while trying to pin down such tables, we are, as second-order plastic structures, not only changing the state we are in, but we are also changing components involved in the determination of our own subsequent structural change. Every machine table could at best be a partially frozen witness of a past situation, and a starting point with an endlessly large amount of possible ways to change. The question would remain thereby: in what state are we now and what will be the following one?
This conclusion cannot be a surprising one. But it makes one thing particularly clear: one cannot participate and observe at the same time. Or differently said: participating and observing are mutually exclusive ways of structural change. This is bound to have consequences for every theory of the human sciences that tries to explain and predict the actual behaviour of human beings. Prediction as a result of a mechanical analysis of actual psychological or sociological behaviour isn’t a realistic goal from this point of view. If Maturana’s approach is accepted, human sciences have to search for another use of their models and theories. Which goal shall we select and why on the basis of this insight? And shall we not have to change our ways of observing ourselves drastically?
F.J. Varela, H.R. Maturana and R. Uribe, 1974, Autopoiesis: The Organization of Living Systems, its Characterization and a Model, BioSystems, vol. 5, North-Holland Publishing Company, Amsterdam.
H.R. Maturana, 1978, Biology of Language: The Epistemology of Reality, in Psychology and Biology of Language and Thought. Essays in Honor of Eric Lenneberg, Academic Press.
A good overview and a critique of functionalism is available in N. Block (ed.), 1981, Readings in the Philosophy of Psychology, Methuen & Co, London.
Found a mistake? Contact corrections/at/cepa.infoDownloaded from http://cepa.info/2742 on 2016-05-22 · Publication curated by Alexander Riegler