Anarchy 31/Anarchism and the cybernetics of self-organising systems
|Anarchism and the|
|cybernetics of self-organising|
|systems||JOHN D. McEWAN|
Firstly, what do we mean by a self-organising system? One definition is simply ‘a system in which the order increases as time passes’, that is, in which the ratio of theexhibited to the maximum possible variety decreases; variety being a measure of the complexity of the system as it appears to an observer, the uncertainty for the observer regarding its behaviour. A system with large variety will have a larger number of possible states than one with smaller variety. Thus such a system may start by exhibiting very varied behaviour, e.g. a large number of different responses to a given stimulus may appear equally likely, but over a period of time the behaviour becomes less erratic, more predictable—fewer and fewer distinct responses to a given stimulus are possible (or, better, have a significantly high probability.)
This definition is, however, in some ways restrictive. The best such a system can do is to reach some sort of optimum state and stay there. Also, if we regard the system as aSuch considerations suggest that it would be more fruitful to incorporate in the definition the idea that the maximum possible variety might also differ at different times. Thus Pask restricts the term to situations where the history of ‘the system’ can best be represented as a series S₀ S₁ . . . Sₙ each term a system with fixed maximum variety, and each self-organising in the first sense. With this definition we are attempting to maintain stability in a fluctuating environment, the types of disturbance with which it can deal are limited by the fixed maximum variety of the system. This point will be dealt with later. The essential thing is that unpredictable disturbances are liable to prove too much for the system.
For an example of self-organising behaviour in this sense, we may consider a human being learning to solve certain types of problem, as his behaviour appears to an observer. Over an interval the behaviour may appear self-organising in the first sense. When, however, the learner adopts a new concept or method, there will be a discontinuity in the development of the behaviour, after which it will again be self-organising in the first sense, for a time, but now incorporating new possibilities, and so on.
In many discussions of control situations the concept of ‘Hierarchy’ appears very quickly. This may tend to make the anarchist recoil, but should not do so, since the usage is a technical one and does not coincide with the use of the term in anarchist criticisms of political organisation.
Firstly, the cybernetician makes a very important distinction between two types of hierarchy, the anatomical and the functional, to use the terminology adopted by Pask. The former is the type exemplified in part by hierarchical social organisation in the normal sense (e.g. ‘tree of command’ structure in industry), that is: there are two (if two levels) actual distinguishable concrete entities involved. The latter refers to the case where there may be only one entity, but there are two or more levels of information structure operating in the system—as for example in some types of  This idea might, I think, be suggestive to anarchists.. A comparable concept is ’s ‘disalienated decision procedure’.
Secondly, even in the case of ‘anatomical hierarchy’, the term only means that parts of the system can be distinguished dealing with different levels of decision making and learning, e.g. some parts may deal directly with the environment, while other parts relate to activity of these first parts, or some parts learn about individual occurrences, while others learn about sequences of individual occurrences, and others again about classes of sequences.
Even in the anatomical sense, then, the term need have none of the connotations of coercive sanctions in a ruler-ruled relationship which are common in other usages.
An important phenomenon in self-organising systems is interaction between the information flowing in the system and the structure of the system. In a complex system this leads to Redundancy of Potential Command—it is impossible to pick out the critical decision-making element, since this will change from one time to another, and depend on the information in the system. It will be evident that this implies that the idea of a hierarchy can have only limited application in such a system.
Prior to this Pask had developed individual teaching machines which were important advances in the growth of applied cybernetics. However, on considering the problem of group teaching (for skills where some calculable measure of the pupils’ performance, the rate of change of which will serve as a suitable indication of learning, exists), he did not simply combine individual machines.
The important insight he had was that a group of human beings, in a learning situation, is itself an evolutionary system, which suggested the idea of the machine as a catalyst, modifying the communication channels in the group, and thus producing different group structures.
In the development of the individual teaching machines, the possibility of the pupil dominating the machine had already arisen. This Pask now extended by introducing the idea of a quality ‘money’ allocated to each member of the group, and used by each of them to ‘buy’ for himself control over the communication structure of the group and over the partial specification of the solution provided by the machine. Now, in the individual machine, the degree to which the pupil was helped was coupled to change of his degree of success. If he was becoming more successful then the help given was decreased. In the group machine, the allocation of ‘money’ is coupled to two conditions—increasing success and increasing variety in the group structure. This second condition is the key to the novelty of the system.
The system, then, has changing dominance and exhibits redundancy of potential command.
In practice, each pupil sits in a little cubicle provided with buttons and indicators for communication, and a computer is used for control, calculating the various measures, etc. The operator is provided with some way of seeing what is going on, and can deliberately make things difficult for the group, by introducing false information into the channels, etc., seeing how the group copes with it.
The problems which Pask, at the time, had used in these group experiments had been formulated as conveying information about the position of a point in some space, with noise in the communication channels. The group had been asked to imagine that they are air traffic controllers, given co-ordinates specifying the position of an aircraft at a certain time, for example.He suggests, however, that problems of agreeing on a choice of policy on a basis of agreed facts is not, in principle, very different from the case in which ‘the facts’ are in dispute, and there is no question of adopting any future policy—except of course the policy to adopt in order to ascertain the true facts and communicate them; this being the problem which the group solves for itself. It is in this sense that
It will be noted that the state of the system when in equilibrium is the solution to the problem. Also that this solution changes with time. This is also the case in the first example from purely human organisation which occurred to me—a jazz band (an example also suggested by Pask).
Pask emphasised that he had not then had the opportunity to obtain sufficient data to make any far-reaching well substantiated generalisations from these experiments. The results he had obtained, however, were very interesting and, I think, give considerable insight into the characteristics of self-organising systems, and their advantages over other types of decision-makers.
Some groups, after an initial stage while they were gaining familiarity with the machine, began assigning specific roles to their members and introducing standard procedures. This led to a drop in efficiency and inability to handle new factors introduced by spurious information, etc. The learning curve rises, flattens, then drops sharply whenever some new element is introduced. The system is now no longer self-organising.
Necessary characteristics for a group to constituted self-organising system, Pask suggests, are avoidance of fixed role-assignments and stereotyped procedures. This is of course tied up with redundancy of potential command.
I think we might sum up ‘fixed role assignment and stereotyped procedures’ in one word—institutionalisation.
Note that these characteristics are necessary, not sufficient—at the very least the group must first of all constitute a system in a meaningful sense; there must be communication between the members, a sufficient structure of information channels and feedback loops.
The role of the computer in Pask’s system may be worrying some. Is this not an analogue of an authoritarian ‘guiding hand’? The answer is, I think, no. It must be remembered that this is an artificial exercise the group is performing. A problem is set by the operator. There is therefore no real situation in actuality for the group to affect and observe the result of their efforts. It is this function of determining and feeding back success/
The other important aspect of the machine as a catalyst in the learning process, we have already mentioned. There is a rough analogy here with the role of ‘influence leader’ in the Hausers’ sense, rather than any authoritarian ‘overseer’. I will return to this question of the role of the machine shortly.
Regarding the group as a decision maker, Pask suggests that this is perhaps the only sense in which ‘two heads are better than one’ is true—if the ‘two heads’ constitute a self-organising system. The clue as to why a number of heads, e.g., notoriously, in committees, often turn out to be much worse than one, is, he suggests, this business of role assignment and stereotyped procedure. He has not, however, suggested why this should arise.Drawing on knowledge of behaviour of a self-organising nature
Imagine a workshop of reasonable size, in which a number of connected processes are going on, and where there is some variation in the factors affecting the work to be taken into account. There is considerable evidence that the workers in such a shop, working as a co-operating group, are able to organise themselves without outside interference, in such a way as to cope efficiently with the job, and show remarkable facility in coping with unforeseeable difficulties and disruptions of normal procedure.
There are two levels of task here:
- The complex of actual production tasks.
- The task of solving the problem of how the group should be organised to perform these first level tasks, and how information about them should be dealt with by the group.
In situations of the kind I am imagining, the organisation of the group is largely determined by the needs of the job, which are fairly obvious to all concerned. There is continual feed-back of information from the job to the group. Any unusual occurrence will force itself on their notice and will be dealt with according to their resources at that time.
Purely for the purpose of illustration, let us now consider the situation of the same type of shop, only this time assuming that it is organised by a committee from outside the shop. The situation in which the committee finds itself is completely different from that of the work group. There are now three levels of problem:
- The problems solved by the individual workers, i.e. their jobs.
- The problem of the organisation of the work group.
- The problem of the organisation of the committee itself.
The determining success/
The committee is denied the continuous feed-back which the group had. While working on its solution to the second level problem, it will have no information about the success of its alternatives, only previous findings, coded, in practice, in an inadequate way. The degree of success will only be observable after a trial period after they have decided on a solution. (Also unusual circumstances can only be dealt with as types of occurrence, since they cannot enumerate all possibilities. This is important in determining the relative efficiency of the two methods of organisation, but is of less importance in our immediate problem.)
In theory, such a controller could still remain an adoptive self-organising system, learning the structure to adopt in particular circumstances over a longer period of time, though it would still suffer from imperfect information.
In practice, however, the committee promptly convene a meeting, assign specific functions and decide on standard procedures. The actual determining information is probably a mixture of personality factors (including externally deprived status) and the existing ideas on organisation theory (including local precedent) possessed by the members. Once decided they will shelve the third level problem unless disaster, or a new superior, strikes, when a similar, but more cumbersome, procedure will be necessary to re-organise the committee along the same general lines.
In other words, within the closed system of the committee and work group, there is no, or virtually no, coupling between the success of the actual undertaking, i.e. the production job, and the decision procedure solving the third level problem. Worse, the factors influencing the solution of this problem, far from increasing the possible variety of the committee, lead to rigidity and low variety. Owing to this structure it will generally prove less efficient than a single imaginative person.
We might suggest, then, that it is this isolation from the process in terms of which the success of their own activity is defined, which is generally typical of the committee situation, which leads to their common failure to exhibit self-organising characteristics, and frequent inadequacy as decision makers.
Consider the first case of the self-organising work group again. Here it is the job itself which provides the analogue of Pask’s machine, as far as feedback of success/
In the lecture referred to earlier, his main thesis was the impossibility of truly efficient control of a complex undertaking by the type of rigid hierarchic organisation with which we are at present familiar. That such systems manage to survive, and work in some sort of manner, as they obviously do, is, he suggested, due to the fact that they are not entirely what they are supposed to be—that there are unofficial self-organising systems and tendencies in the organisation which are essential to its survival.
Beer is unusually perceptive, and frank, in emphasising the prevalence and importance of unofficial initiatives at all levels, e.g. (of shop-floor workers). “They arrange things which would horrify management, if they ever found out”, (of charge-hands, etc.) “If they did not talk things over and come to mutual agreements, the whole business would collapse.”
The main keystones in Beer’s argument are’s ‘ ’ from the theory of , and information-theoretic requirements for adequate channel capacity in a multi-level system.
The principle of requisite variety states that, if stability is to be attained, the variety of the controlling system must be at least as great as the variety of the system to be controlled. We have already had an instance of this, for this was really the trouble with our hypothetical committee: due to its rigid structure and the need to issue instructions in terms of standard procedures to be adopted, it could not possibly be efficient in a situation of any complexity. If we made the further assumption that there was no organisation of the work group other than that imposed by the committee, chaos would be unavoidable. Approximations to this occur in ‘’. In normal working, the initiatives of the shop-floor workers would serve as an additional source of variety, this enabling the principle of requisite variety to be satisfied, at least as far as normal variations in the factors affecting the production situation were concerned.
The relevance of the requirements of channel capacity is to the inadequate, attenuated information available at the top of the hierarchy—this is inevitable, for, in practice, the channel capacity could never be made adequate in the sort of pyramidical structures we have—and also to the inadequacy of the formal channels between subsystems (e.g. departments) which require to co-ordinate their activities.To emphasise how far conventional managerial ideas of organisation are from satisfying the principle of requisite variety, Beer used an
In discussing the attempts of an inadequate control system to control a system of greater variety, Beer pointed to the accumulation of unassimilable information likely to occur as the control vainly struggles to keep track of the situation.
A comparable converse phenomenon was pointed out by  (The first italics are mine.)in 1851, in what must rank as one of the most prophetic statements about the development of social organisation ever written: “(The government) must make as many laws as it finds interests, and, as interests are innumerable, relations arising from one another multiply to infinity, and antagonism is endless, lawmaking must go on without stopping. Laws, decrees, ordinances, resolutions, will fall like hail upon the unfortunate people. After a time the political ground will be covered by a layer of paper, which the geologists will put down among the vicissitudes of the earth as the papyraceous formation.”
This is also an early, and lucid, statement of the complexity of the control situation in social organisation.
Beer has some suggestive ideas on the question of centralisation vs. decentralisation in industry. (That is, centralisation of control. The question of centralisation of plant is a different, if related, problem.) He puts the dilemma thus:
The point, he suggests is that neither alternative corresponds to what we find in really efficient systems, i.e. complex living organisms. What we do find are a number of different, interlocking control systems. Beer also draws attention to the prevalence, and importance, of redundancy of potential command in self-organising systems, and points out that it is completely alien to the sort of theory of organisation found in industry and in similar undertakings.
The type of organisation at which we should aim is, he suggests, an organic one, involving interlocking control systems, intermeshing at all levels, utilising the principle of evolving self-organising systems, with the channel capacity and flow of information kept as high as possible.He mentioned in this connection an American businessman who claimed that his business was, in part, organised along somewhat similar lines and seemed to work very well. The idea was that anybody at all, no matter how ‘junior’ (I do not know whether this was actually restricted to what are termed ‘staff’ or not), could call a conference at short notice, to discuss anything they wanted, whether connected with
In context of interlocking control structures, we may note, as a fairly crude example, theattempt to co-ordinate the activity of their basic units, the factory unions, through an interlocking two-fold structure of industrial and territorial federation.
Let us now contrast two models of decision making and control. First we have the model current among management theorists in industry, with its counterpart in conventional thinking about government in society as a whole. This is the model of a rigid pyramidal hierarchy, with lines of ‘communication and command’ running from the top to the bottom of the pyramid. There is fixed delineation of responsibility, each element has a specified role, and the procedures to be followed at any level are determined within fairly narrow limits, and may only be changed by decisions of elements higher in the hierarchy. The role of the top group of the hierarchy is sometimes supposed to be comparable to the ‘brain’ of the system.
The other model is from the cybernetics of evolving self-organising systems. Here we have a system of large variety, sufficient to cope with a complex unpredictable environment. Its characteristics are changing structure, modifying itself under continual feedback from the environment, exhibiting redundancy of potential command, and involving complex interlocking control structures. Learning and decision-making are distributed throughout the system, denser perhaps in some areas than others.
Has any social thinker thought of social organisation, actual or possible, in terms comparable with this model? I think so. Compare Kropotkin on that society which “seeks the fullest development of free association in all its aspects, in all possible degrees, for all conceivable purposes: an ever-changing association bearing in itself the elements of its own duration, and taking on the forms which at any moment best correspond to the manifold endeavours of all.”
Further, “A society to which pre-established forms crystallised by law, are repugnant, which looks for harmony in an ever-changing and fugitive equilibrium between a multitude of varied forces and influences of every kind, following their own course.”
The language is perhaps somewhat vague and ambiguous, but for a brief description in non-technical terms, of a society conceived as a complex evolving self-organising system, it could hardly be bettered. Certainly not in 1896.
The tragedy is not that so-called progressive thinkers today think that anarchist ideas of society and social organisation are inadequate. (This is excusable, and indicates failure on the part of anarchist propagandists to develop and spread their ideas.) It is that they think the other model is adequate. Also that they are incapable of thinking in any other terms.Hence such thinkers are surprised when they cannot find the great efficient decision makers they expect in control of our institutions. The
Again, when existing systems break down, as in industrial disputes, the tendency for the leaders on both sides is to attempt to remedy the situation by measures which increase the inadequacy of the system. That is, they attempt, by reorganisation and contractual measures, to increase the rigidity of the system by defining roles and responsibilities more closely, and try to confine the activities of human beings, who are themselves evolving self-organising systems, within a predetermined contractual framework. An interesting example of this will be found inby .
To return to the conventional picture of government and the supposed control by the governed in democratic theory:
Firstly, does what I have said about the inefficiency and crudity of the governmental model as a control mechanism conflict with Grey Walter’s analysis in his article “The Development and Significance of Cybernetics” in Anarchy 25, in which he claimed that Western democratic systems were remarkably sophisticated from the cybernetic point of view?
I do not think so. The point is that what I am claiming is that they are inadequate for controlling the economy, say, or providing the greatest compatible satisfactions for the governed, as Proudhon pointed out. I would also claim that they are inadequate as mechanisms for maintaining order in society, unless society is conceived as largely self-regulating without the governmental institutions. Given this, I do not deny that the government-electorate system has proved an efficient machine for maintaining itself, although I might be inclined to give a little more importance to unofficial, informal elements in the system in this context than Grey Walter does in his article.
I agree that the system is well adapted to this task. Also, various psychological factors outside the scope of cybernetics help in the self-perpetuation of a system of this nature.
If the model of effective control by the government is inadequate, the naive democratic theory of control of the government by the people is much more so. This theory puts great stress on the importance of elections as the means by which the governed control their rulers and on the results of the elections, and hence, derivatively, on the constitution and behaviour of the government, as expressions of ‘the will of the people’.
If we consider the indiviudal, in a two party system, he is allowed one binary choice every five years or so, in which to reflect all the complex, dimly understood, effects of government actions, intended and unintended. The model seems to allow of no structured subsystem to be identified as ‘the people’—there is only an aggregate of individual choices.It seems to me significant that this theory of self-government of the people, by the people, through universal, or at least wide, suffrage,
This is not to deny the genius of some of the thinkers who worked within the limitations of this model of democracy, for they were able to see the difficulties in practice, and devised most complicated systems of checks and balances to render their systems practicable, (e.g. the architects of the American constitution, as Grey Walter points out). However, they could not be expected to overcome the fundamental inadequacies of their model of government of the people, by the people, for the people, no matter how successful they were in developing the skeletons of viable self-perpetuating systems.
In contrast to the ‘rabble hypothesis’, we find that libertarian socialist thought, especially in Kropotkin and Landauer, showed an early grasp of the complex group structure of society; society as a complex network of changing relationships, involving many structures of correlated activity and mutual aid, independent of authoritarian coercion. It was against this background that they developed their theories of social organisation.
Neither am I convinced by the more sophisticated pressure group theory of democracy, introduced in an attempt to avoid the obvious inadequacy of the naive theory. As a descriptive theory of the actual situation it does seem reasonably adequate, but as a means by which the individual obtains a voice in decisions affecting him, it is just as inadequate as the naive theory. This in fact is generally admitted by its adherents, who have largely dropped the idea of democracy as self-government.
In the case where a group, of a self-organising type, freely organises itself to tackle some situation, the resulting structure adopted by the group might be taken to exhibit ‘the will of the group’. More generally, groups of this nature are capable of genuine group decisions. Such expressions as ‘the will of the group (people)’ are, I suggest, acceptable, and only as a rather dangerous shorthand, solely in cases of this sort.
In direct application, this is, of course, limited to fairly small groups, since, beyond a certain size, an unstructured aggregate of human beings is unable to act as a group, because there is too much information to be handled. The channel capacity is probably inadequate, and, even if the individual member could be presented with sufficient information, he would be unable to deal with it.In certain work situations where the job effectively constrains the system, and only part of the behaviour needs to be correlated, we might expect larger aggregates to be capable of behaviour as a group. This is borne out by experience. In a situation where complex activity has to be correlated and there are few prior constraints, e.g. collective improvisation in a jazz band, most research groups, discussion groups, a maximum of the order of ten seems to be imposed; in manual jobs
We have said that only small aggregates of human beings, if regarded initially as unstructured, can exhibit genuine group behaviour. There is no reason, however, why large aggregates, if sufficiently structured, should not maintain coherent behaviour, while retaining genuine self-organising characteristics enabling them to deal with unpredictable disturbances in their environment (including in ‘environment’ their own ‘substance’, i.e. the human being constituting the aggregate) without developing a hierarchic structure in the authoritarian sense.
This is not to say that there will be no hierarchy in the logical sense. There will certainly be functional hierarchy in the sense of multi-level information flow, i.e. problem solving at the level of group environment, internal activity of subgroup, relations between subgroups, and so on. We have seen that this need not necessarily mean different isolatable physical parts handling the different levels. In a situation of great complexity, however, we would expect to find anatomical hierarchies, in as far as there would be identifiable subgroups, of varying degrees of permanence of form and constitution, dealing with different levels of activity.
The essential points are that the existence of redundancy of potential command, with changing dominance, means that any analysis of part of the system at any time in terms of a hierarchic model must be regarded with caution, and that, where such anatomical hierarchy is distinguishable, it need not be a question of the higher levels controlling the lower by coercive sanctions, but rather of feeding back information to bias the autonomous activity of the other subgroup. In short, a very different sort of hierarchy from that of managerial theory.
There certainly need not be any isolatable ‘control unit’ controlling the rest.
I am using ‘structured’ here in a sense comparable to Buber, i.e. possessing a structure of connected subgroups, groupings or subgroups, etc., of a functional nature, but I would place relatively less emphasis on formal federation of subgroups, even in multiple federation, than Buber, and more on more complex forms of connection. Also I am counting as subgroups both localised and more diffuse structures, formal and informal. One form of connection which seems to be of importance, is the case of diffuse substructures ‘penetrating’ into more localised ones, e.g. certain members of a particular subgrouping being members of some more widespread grouping, some sort of interest association, say, and thus serving as a means by which information about special forms of activity, passing in the more widespread structures, can pass into the localised structure, and play a part in determining its subsequent behaviour.
This, however, is no excuse for remaining bound by a primitive and inadequate model of decision-making and control procedures. The basic premise of the governmentalist—namely, that any society must incorporate some mechanism for overall control—is certainly true, if we use ‘control’ in the sense of ‘maintain a large number of critical variables within limits of toleration.’ Indeed, the statement is virtually a tautology, since if such a situation did not exist, the aggregate would not possess sufficient stability to merit the designation ‘a society’.
The error of the governmentalist is to think that ‘incorporate some mechanism for control’ is always equivalent to ‘include a fixed isolatable control unit to which the rest, i.e. the majority, of the system is subservient’. This may be an adequate interpretation in the case of a model railway system, but not for a human society.
The alternative model is complex, and changing in its search for stability in the face of unpredictable disturbances—and much less easy to describe. Indeed, we are perhaps just beginning to develop an adequate language to describe such situations, despite the prophetic insights of a few men in the past.
A quotation from Proudhon makes a fitting conclusion—and starting point—“People like simple ideas and are right to like them. Unfortunately, the simplicity they seek is only to be found in elementary things; and the world, society, and man are made up of insoluble problems, contrary principles, and conflicting forces. Organism means complication, and multiplicity means contradiction, opposition, independence.”
- See : Decision-Making and Productivity (Blackwell, 1958).
- : “Interaction between a Group of Subjects and an Adaptive Automaton to produce a Self-Organising System for Decision-Making” in the symposium Self-Organising Systems, 1962, ed. , and (Spartan Books).
- See : Cybernetics and Management (English Universities Press, 1959) pp.123-127, and : An Approach to Cybernetics (Hutchinson, 1961).
- See and : The Fraternal Society (Bodley Head, 1962).
- See, for example, the paper by Geoffrey Ostergaard in ANARCHY 2. Note the appearance of new elements of job rotation.
Despite his emphasis on the formal aspects of worker organisation, ’s analysis (see Note 1) of the worker decision process at brings out many of the caracteristics of a self-organising system: the evolving nature of the process; the difficulty of determining where a particular decision was made; changing dominance; the way in which the cumulative experience of the group changes the frame of reference against which subsequent problems are set for solution. A better idea of the gang system from which this derives can, however, be obtained from Reg Wright’s articles in ANARCHY 2 & 8. on collective contract working in the coalfield quoted by in A New Approach to Industrial Democracy (Blackwell, 1960) and the discussion of this book by
- : (Freedom Press, 1923).
- Compare also the concluding section of ’s An Approach to Cybernetics, in particular the discussion of a ‘biologically organised’ factory.
- Peter Kropotkin: (Freedom Press, 1895).
- See, for example : (Penguin, 1954), Ch. 2.
- See G. Ostergaard’s discussion in ANARCHY 2. : A New Approach to Industrial Democracy and
- See Martin Buber: (Routledge, 1949).
- Buber : The Theory of Taxation (1861) quoted in