The interdisciplinary and intersectoral round-table discussion took place as one of the working groups of the preliminary meeting on "Exploring the Network alternative" which was held at the University of Concordia, Montreal, Quebec, Canada, 18-20 November 1976 (see report on pages 352-3551.
The purpose of the discussion was to examine the problem of clarifying the concepts associated with "network".
The orientation was provided, in part, by background documents which drew attention to the increasing use of " network" in connection with "peoplegroups" and certain kinds of interorganizational activity, as distinct from various conventional uses such as in «social networks" of individuals (which are a special preoccupation of a certain school of sociology). Orientation was also provided by informal presentations from Or M. Vidyasagar (electrical networks and grids), Or R. M. Chen (automation control circuits), Or P. Oansereau (biological and ecological systems), and Or Joseph Fiksel (networks as mathematical objects). A case history of a network of individuals and groups based in New England was presented by Marc Sarkady (Another Place), complemented by a presentation by Linda LeClair (American Friends Service Committee) on one of that network's particular concerns, namely to campaign against the establishment of a nuclear plant in the area.
The extracts presented here locus on the main theme of the discussion during the round-table: as to whether there was any real distinction between a " system" and a " network" and. if so, what that distinction was, particularly in the case of networks of individuals, groups and organizations. (This point was also explored in the group on « Complexity" during the meeting of the International Foundation for Social Innovation, Paris, March 1977. The summary report by R. P. Dubarle is included in this issue, pages 369372).
The transcripts have been edited by one of the participants, Anthony Judge, I without submitting a draft to the original group. Speakers are not identified .by name for this reason (and also because not all speakers could be identified from the tapes). Known speakers are identified by bold letters A, B, C, 0, E, F, G and H. Unidentified speakers are labelled by the letters X, Y or Z, as appropriate. The editor is however responsible for the final wording since discussion of related themes has been omitted as well as detailed exploration of some points. The original participants, and others, will be invited to make further comments in the light of this report.
It may be asked why it is necessary to adopt the space-consuming method of reproducing the (almost) verbatim statements made during the discussion, rather than a synopsis, which would normally be adequate. The answer is that it is in fact the original statements which best reflect the current "informed confusion ", the relevance of unexpected insights, the avenues which could be explored, as well as the provisional nature of any conclusions at this stage.
It may also be asked, as was done during the discussion, whether the differences, if any, are significant, rather than simply a matter of terminological preferences, current fads, or plain quibbling. Specialists from some disciplines may be quick to reply.
Those reading this report of a discussion between specialists from different disciplines, together with others interested in the social significance of such a distinction, may recognize that the matter is not so simple.
The issue itself is however very simple. Terms such as " international system", the " establishment system", or « the System" are widely used, whether by social activists, academics, or politicians. For some they have extremely negative connotations, and such people increasingly prefer to think in terms of "networks" and "networking ", which for them represents a distinct method of organization (or minimal organization). Are they deluded and misinformed, or is there a real distinction between working in (or with) systems and working in (or with) networks?
A It seems to me that we're not struggling with two different things - content and structure - but with at least three different levels. We start with the assumption that there is something we might call a network or a network structure in various fields - in electrical phenomena, in political systems, and so forth. Each of those networks carries different kinds of messages carries different content. At one level we can set aside whether it's a pronuclear network or an anti-nuclear network. The content of it is one level. Another is the structure of the network; what are the characteristics of structure how do they differ - morphology. Then there is a third level which we keep sliding in and out of and that IS the epistemological questions that the whole thing raises. For example if in fact you have a vocabulary that you can use in electrical terms, can you transfer that to biology, or social networks without a whole set of assumed analogies - and so on. So it seems to me there are questions of language, semantics and epistemology at one level. There are questions of morphology at another level, and there are a whole set of questions about the character or substance of the network at another level.
My own interest is not in the content, so called, but in the structure, and also epistemology. We had Dr. Vidyasagar give us a kind of elementary lecture yesterday on the terms used in electrical networks, and to me it was interesting to see how well elaborated the vocabulary is. I would like to know what the parallel - if there are parallel terms - would be for biology, for ecosystems, for political systems. Might there in fact be parallel vocabularies for each of these fields, all of them saying different things? Perhaps I should say here that as a writer, a student of social change, interested in emerging political institutions and so on, my concerns are with analogies I might be able to fetch or borrow out of this discussion we have today ...
B Could I come back to this problem of network or system - if one directs the way the network functions is it still a network? I'm interested in the flip. You chose system in your example, is this system embedded in a network? If we take the example of the anti-nuclear campaign, they were concerned with a system to handle that issue. The system was still embedded in a network and when the issue disappears the system will disappear as a system. The network may well remain.
C I would think the reverse. A system as I see it is definable in terms of agentry, source, process, trophic level. And this is unique to a particular trophic system. That same combination of elements does not repeat itself but goes through a series of variations. Within each ecosystem there is of course a network. Whether the network is contained in an ecostystem or extends beyond it, or whether it is autoregulated or dependent and feeling out to other ecosystems is very relevant. To me network is a sort of «in-nervation" of one or more ecosystems. It 'may be two ecosystems will be so interdependent and the loops between them so small that that network will appear to have more cohesion than either of the two ecosystems, especially if they are indeed absolutely interdependent.
B Can we focus on a terminology or glossary? It seems there are reference books we could use to build this. There are some critical areas of terminology, such as " node". I would be happy as a fi rst round of a glossary to put down about 150 concepts and at another meeting to go through and decide on their relative usefulness.
A I don't think it is a question of which are more important and which are less important. I think you could map which concepts recur from field to field and which concepts are discipline- specific, then you could question the difference. I was thinking of more than one column, biological terms, mathematical terms, political terms, etc.
B I like the suggestion of «columns" and perhaps another column to suggest social implications. For example " maturity" of an ecosystem or climax we don't have a feel for this in social networks, but it is well defined in ecosystems. What interests me in the notion of maturity is that there is a measure of how mature the system is which is in fact related to the number of " links" between species at different trophic levels.
C A mature ecosystem is homeostatic and very largely self-regulated or else has long been associated with other ecosystems in a self-regulating complex.
B We jump immediately to the conclusion that this is what we would aim for in a "limits to growth" idea. If in ecology there is a sense of what kind of system this would be, one could gain one or two insights as to how it might be in a social system.
C I think what we are aiming for is better and more far seeing controls that will have to manage even more of the existing ecosystems of the planet.
B The interesting thing is " who manages" and whether it is still possible in this society to imagine some unique central controller who is managing the planet through various subsidiary controllers. This is no longer a viable
C Natural ecosystems are harmonious but this harmony and stability, maturity if you want to call it that, has been achieved at the expense of efficiency.
All natural ecosystems are relatively, rather inefficient. Take the Sequoian forest, it is doing a very poor job. Of course take down the Sequoian forest and put in Douglas fir and you'll get a tremendous crop in no time at all.
A But I think the point is still valid. If the measure of efficiency is how many Sequoia you grow, it would be a very efficient system. Well, what I've heard resonating in the background of this conversation is the idea of maturity, the idea of mature networks. What I begin to think is maybe one can view a social system as a process of « growing networks» and that one can even think in the form of «social agriculture ·, in which you are "artificially fertilizing" the society so that it produces certain kinds of networks. That's just the imposition of an agricultural analogy on sociology.
B I find that very interesting. I'm still troubled by this problem of network or system. Do the systems fit into the network or do networks fit into the system? I think it a problem because we run the risk of people saying well you can't use this word "network» when basically it's just a " system ».
D It seems to me that it's a choice of model If you want to focus on loops. on exchanges then you're thinking about networks. If you want to focus on boundaries and demarcations then you think about systems. The other thing I notice is that people who think about systems seem to like to have a box that has one purpose. where as people who think about networks have a feeling the network may have many purposes. There are two ways of getting things done, you can set a purpose and then bring people together to work on that purpose, or you can go around looking for interesting people and bring them together and pat them on the back. In both cases you will have something happen, but in the one case you don't know what's going to happen you have a feeling it will probably be good, because you've had some discrimination in whom you have brought together.
X A network is a means. A network is exploited by a system.
B You Iike the word network and I wonder how you see that in terms of people talking about establishment systems.
A I think that what he says is so and the list of suggested differences (see page 365) are essentially connotations.
B Well this has consequences that if you want to build a systems model you are quick to define boundaries.
A You say the «systems» model. Is it a different model or is it simply a different vocabulary for describing the same kinds of relationships? Is there something that defines it ? I think it's a valid issue for us to focus on. I need to be convinced that there's a real life difference, as distinct from a set of linguistic preferences, but I do not deny that there might well be. I want to be educated on that.
X A system can describe an organization or a manufacturing process. Can a network do the same?
C May I volunteer a definition? A network consists of strands that ensure the cohesion of the system and / or allows it to tie up with other systems.
D It may be that or a network may cross many systems.
C A network is the «innervation" of the system. It ensures circulation within the system and occasionally ties up with other systems.
A Bell Telephone and Bell Canada use both terms to refer to the Bell " system" and the «telephone network". Now the Bell system is a social, economic and political organization. The network is a physical structure to them. That may be because the entire company is run by engineers.
D And it is a North American «telephone network». The Bell «system» manages a component of the North American telephone network.
B I find that an interesting contrast. Because then for me the network is larger than the system. The system is a way of controling part of a network.
A But they could be co-extensive.
X Or they could be the other way around.
A One big telephone company for all of North Americans and several in Canada. But consider the question that was asked over here about a manufacturing system as producing an end product. Can a network work that way too? Maybe that will lead us somewhere. It has to do with time - with networks through time.
X Can I ask why it is so important to differentiate between a system and a network? Should the difficulty not tell us something?
B I think the reason why it is important lies in the fact that there is a whole literature and many organizations concerned with systems of one kind or another. And it seems to me that there is a whole other area of concern with networks. And the people concerned with networks are not necessarily served by, or interested in, what the people concerned with systems are doing.
A I think there is a real political difference in the two words.
D The unipurpose thing I think very important.
A And I think people react rather strongly in different ways to the two terms. We talk about system and they think the thing is imposing itself on them, whereas being part of a network that's always sexier.
X It may change with time.
A It might, but it seems that is the present political loading.
X What about exploiting the technology of systems then. A system does something, never mind the political implications. System has a purpose and a methodology. I think surely they overlap in many ways.
B I think that the advantage with networks though is that they are multipurpose. They are prepared to respond to a variety of conditions and problems.
X Systems can too.
B Only if we define them in advance, and this is the problem.
D Is it not a problem of connotation and emphasis?
X Again, it depends on who is describing the system and for what.
A I think we should spend more time on this as it really is the background to anything else we might discuss. There is so much terminology linked to systems and if there is a distinction we have to make it. The telephone is a democratic technology as distinct from, say, broadcasting where you have a centralized information factory which pumps out an Image. A telephone system --- one of the problems they had was by law they had to provide service - they have to give it to anyone who wants it. They can't make political or other kinds of distinctions. What that means is they have no control over demand. The system is activated by the consumer.
X I would think there would be something more interdisciplinary about a network.
C Should we ask ourselves whether it is at all possible for a system to exist without having some kind of network as an inner structure? My answer would be no. Any system has some kind of an interior network. That is a minimal proposition as far as I am concerned. Now whether the network does and can extend to more than one system, the answer is of course yes it can. In fact, there would be very few systems that had a very self contained network not extending beyond their boundaries. This is almost unimaginable and certainly very exceptional.
A But what is the distinction?
C To me the network is some kind of conveyor of whatever is operative within the system; what I would call in my vocabulary resources, however understood, whether it's oil or information. It is still a resource and in order to pass from one state to another, to be transported and transformed by some kind of energy, it has to borrow the pathways of a network. We were considering hierarchies this morning and I think there are hierarchies in a system, and there are bound to be hierarchies in a network. Some channels within a network are all-purpose conveyors and other channels are highly specialized and carry only one kind of information.
E For me a system is more defined and constrained than a network. Think of the distinction between a heuristic and an algorithmic process with certain restricted inputs and predictable restricted outputs. Heuristics are processes which operate on a whole class of things with only partially predictable results. This may just be my yearning for a network to be very flexible, so that you never know for sure what's going to happen. I tend to think of a system as something predictable.
C It's predictable in as much as you know what is being carried and what mechanisms will stop or will forward the resource. There are all kinds of signals that say go and stop, or on and off. A network is full of on and off signals, whether you're dealing with a bulbous plant that has a rest period, or whether you're dealing with a bank that has opening and closing hours. There is a stop and go - now what activates the stop and go ?
E Well I would talk about the banking system, because banks are among the most predictable, in their mode of operation.
A I suspect that you would find many bankers who would be very unhappy with your describing the system as predictable, when the money system ...
E Yes, but I think that they're looking at a different system than I am. As a user of banks, I am looking at the surface manifestations of branches and transactions and things of that kind; so I think we're looking at different systems.
X You're looking at the mechanics. You car describe the mechanics and interconnections quite precisely because you build them.
A You don't build a solar system.
F And you wouldn't call a solar system a network.
A Well, OK, maybe that's a point that we can work on.
B I think that the predictability in something like a solar system, or an ecosystem, derives more from what we want to identify.
A Is that predictability, what we call predictability, isn't that really a reflection of a level of human mind, rather than some objective fact - until Newton we couldn't predict very well...
C There might be some natural factors - the predictability is high where the freedom is low. If there are few possible alternatives for a resource to engage in one or more circuits, then the predictability is high.
G It seems that you might be able to say that a system is a subset of a network, but that you might not be able to say it the other, way around.
A We said it the other way around this morning - but I'm not sure. that anyone is persuaded of either of those alternatives. As we said before, you wouldn't call a solar system a network. It seems to me a good pace to begin to look for the distinction. Is the distinction simply a matter of predictability. It seems to me that it's always the human being that's doing the prediction, nobody else is doing it. How well we do that or how poorly we do that varies.
X Doesn't that fail for lack of interconnections - channels and nodes -' channels of information flow within the solar system.
B Let's take that a step further. Assuming that we have an interplanetary society, in which communications were taking place between all the different planets. Those communications would in effect, as they patterned themselves, constitute a network, which would have a much more unpredictable component than the solar system as we know it now.
G Is it the difference between open and closed systems? That an open system is a network, and that a closed system we would not want to characterise as a network.
B I agree; but what bothers me there is that we're defining a network in terms of a system, and I'm not sure that it shouldn't be the other way around. When you said in your opening statement that a system has network as an inner structure, the network is inside the system; and then you add to that that few systems have self-contained networks ..
C The system can be coherent, I think that was the term that I used, in as much as it has a network that irrigates its different parts, that allows it the distribution of resources that makes it what it is - whether it's a bog or a bank makes no difference. The network must in some cases extend outside of the system. I think that a better example of the distinction we're looking for is the surface communication network. We take a map and there is a road, a network of surface communication; but it pervades how many systems - wilderness, farmland, industrial development and urban,
F Could one possible distinction be that the system could be goal-directed, whereas I don't think, at least so far, that a network can be.
D Well, the emphasis is there. Usually when you try to design a system, you try and establish a hierarchy of goals, so that the system won't be conflicted within itself. A telephone network for instance can be used for thousands of purposes.
F ... so that it doesn't have a purpose or a goal built into it, whereas a system does.
D Or it may have the meta-goal of bringing people into communication; or there may even be a meta-meta goal of sustaining hope (which I suspect may be behind a lot of these systems here).
B What about the distinction we were looking at this morning between the Bell «system" and the «telephone network". There in a sense the Bell system is managing the telephone network.
C That really could be an error in vocabulary, because something or other calls itself the Bell system. I would go so far as to say that rarely, if ever, is a network co-extensive with a system.
A I still don't see the difference.
D It's two different ways of modelling things. In system modelling you think a lot about boundaries and boundary conditions and about teleology and purpose, and the purpose of the components; and in the network you think a lot about what properties you're going to concentrate at nodes, and what kind of flows you're dealing with, and the rates of flows.
A But still they could in the end be identical. You could go into the process in one sense in the case of the system, looking for a goal or purpose, etc.; and into the network shooting for other things. There is no mutual exclusivity. You could come out with two things that look alike.
D If you improved your system model you might find that you'd have to use a model that had a network configuration. Or as you improved your network model, you might find that parts of the network had a lot of integrity and that they were mostly connected with themselves, and not much connected elsewhere. So you might as well put a wall around them and call them a system.
C The important distinction to my mind is that the system generates resources, including information. The network does not generate anything, it conveys, that's all it does.
A But if all a network does is convey, passively, what about human networks, or social networks, where each of the nodes - each individual as a node in the network - transforms the information flowing through it ?
C No, it is not the network itself that transforms. The system is a matrix, and the network is the innvervation thereof. So, at a certain point in the network. the network reaches to an area where, for instance, vegetable is transformed into animal material. There is a threshold which is crossed, a transformation that is effected; and a channel exists or does not exist or is blocked. or is on, or is off, which conveys this new information from one node to another.
A But translate that into human beings. If we're talking about a social or political network (we were talking about the experiments with message deterioration); or we have a political party. The political party is a network. It's not a party, it's a network. At one end somebody says, « Let's stop the building of this nuclear plant in New England ", and that filters through the network. By the time it comes out the other end it may say «Let's help build that plant! "
C It is not the network that generates the information. It is the President of the Aluminium Company who says he wants to close the plant; this decision is made in the Executive Offices. He picks up the telephone and this already generated- information is, through the wires, transferred as a message that the plant must be closed.
A You're still conceding in that case the wires of the telephone system as the network. I'm saying forget that. Let us take all the technology out of it. We've got a group of people who form an invisible college, who are interested in ecology before anybody else is interested in ecology. They telephone each other, and so on. Now in loose parlance we call them an ecology network. What's happening though is that as the information moves, through that system ...
B Why did you say « system» ?
A OK... Well, the reason is that I'm still not convinced that there's a distinction - I'm still using the two terms tentatively, interchangeably, until I can find a good distinction. But what I am saying is that if we all leave this room, we would be a network on networks, whether we like it or not a meta-net. .. Assuming that we don't use any special communications technology to communicate, what makes us different from a system. We decide that a network is a such-and-such; and then we have friends out there who are also interested in networks and so forth. By the time I go home and tell my wife what we talked about here. I will not deliver it to her in the form in which I received it. I will re-generate that information. It passes through me and comes out different. So that in that sense the network is not just a passive conveyor, in the sense that a telephone network is a passive conveyor.
C The system is the community of presumably knowledgeable people who have gathered together. That is the system. The matrix is the round table. Suppose that we come out with a definition that we think is pretty good, and we decide that this has to be communicated to Professor so-and-so. We then borrow another network, using the telephone, to convey this same information.
A But in fact the message that passes through this node would not come out the way it went in. What I'm saying is that, even as in a human network, it's quite clear that if the individual humans are nodes in that network, those nodes are quite capable of generating, or re-generating, or altering the message; so are certain mechanical systems capable of doing that - of transforming the information that comes into them into another form.
G But only in certain ways. When you talked about the solar system as being predictable, what you were saying is that we can describe the way in which the parts are related. When we talk about the circulatory system, we can describe the way in which it operates, and for all intents and purposes we assume that all circulatory systems of all human bodies operate like that. And in that sense it's a static model - we can draw a picture of all circulatory systems for all people, and it's going to be the same for each one. But in a network we could never draw a picture of a network that would fit for all cases; so a network is dynamic in a way that a circulatory system is not.
B I feel the lack of inputs from other areas, from medicine, from topology or some specialized branch of mathematics. I can list 10 or 20 such areas which I would like to feed into this process so that we can just see what kinds of concepts we might usefully deal with. Work is required and that will be done in some sort of preparatory document for a future meeting or as a follow- up to this meeting. But what I would like is some feedback on how you feel about where we might go from here with respect to the terminology in question.
H I'd be happy to give you some input as a mathematician concerned with networks.
A I was just going to say. My own feeling is that the time is valuable and I would rather we spent the time on substantive discussion than on procedure for constructing the next meeting.
H Traditionally, networks have been thought of as collections of nodes joined by lines Now when people talk about a system they often draw diagrams that look very similar to this. They simply draw boxes with arrows between them. Now I trunk that IS the basic difference. There is a difference In the notion of the system where you have, say, various boxes with flows be>tween them and so on, and the network. In the sense that the network is a representation of some substantial material object whereas a system represents process or abstractions. In other words, each of the boxes in the system represents something that is going on, perhaps some kind of communication or a creative process, or some process of combining elements or separating elements, or whatever. In a network the individual nodes actually represent entities or positions. It's a more fundamental kind of representation and I personally have worked much more with this sort of representation than with systems.
To put it very simply, the nodes represent simple objects or concepts or whatever. But if they represent concepts, that departs from their being something that exists, something tangible.
Let me give you an example of a concept. For instance, one field that I've worked in is in modelling human memory in which I used nodes to represent concepts such as «warm». But what is it ? It's a simple concept. It's not a conglomeration of ideas. It's not a process or a set of equations or a transformation. A node is just a thing we can some how deal with, that we can separate as an entity, and we can also postulate that this thing has a relationship to other entities. Anyway a system is definitely a more complex notion. It requires more definition of what's going on and it usually involves a flow. a dynamic process. A network is much simpler. It exists in and of itself, even if a network isn't doing anything, it's still there.
X How does a network do this?
H At this point it might be useful to talk about particular kinds of networks because we are talking about a mathematical object which is very general and networks can be applied to lots of situations. Let's take a biological example. If you take a micro-organism. you can think of the cells of the micro-organism as being a network because they are very closely connected. Each cell does communicate with its neighbours. So you can think of this as a communicational network. Now this network definitely does something. It somehow manages to grow. to propagate, to feed itself, and so on.
So networks do things in particular contexts. If you can conceptualize some material object or some organism as a network, then you can talk about what it does, and the interesting thing is trying to figure out how it does that given its structure.
B We have different concepts of network, control networks for example. Here you are talking about a biological example, you are talking about a way a network grows. Now I've been interested in the distinction between the way a network grows according to a fixed pattern and the way a network might evolve. In other words how it adopts a new pattern, a distinct pattern.
H Have any of you ever heard of the mathematical game called « life .. ? It's really a kind of a process. What he does is he divides a plane surface into a grid pattern, and then he puts some entities into some of the cells. These represent things that exist and that grow. They have the power to propagate or they will also vanish under certain conditions. What he does is he goes through steps, one step, after another and at each step some of these entities will (according to these rules) either appear or disappear in the cells. So if you have an evolving network, if you are evolving according to very simple rules, he has shown that these networks can expand, they can contract, they can throw out branches, they can break up into distinct networks which actually move away from each other, and so on. They do all kinds of amazing things depending on their shape at anyone time and the distribution of the entities. For instance starting from an initial shape, the system will go through a number of transformations. It may expand and then contract, and eventually it may reach a stable point where it just oscillates, it can do a flip-flop kind of thing. There are a lot of people now working with this sort of network, and seeing how they evolve different patterns.
X You know you just used the word " system». Could you go back to the biological cell and show the network of the cell. One back. You've described the network of the cell, how is the cell a system ?
H Now you can think of an individual cell as a system in the sense that it does certain things. A system generally has inputs and outputs and transformations upon them, so the cell takes in food. takes in oxygen, it creates certain proteins and so on. It does a certain amount of work and also sends off some products tor other cells.
E I was going to say that's exactly what the system does.
X I don't see any difference between that description that you've just given and the network, at least in the case of a human / social network. You could visualise a human being as a system as a whole, or as a network of smaller components, each of which is a little system.
A Well, the distinction they think you were trying to make was that in one case you said that the network is there (whether it's doing anything or not) and the system is doing something. I would argue in the case of any living network it is doing something whether we value that production or not, whether we call it production, whether we call it doing something. The very act of being is doing something.
H There is no argument here. I just talked about a network as an abstract concept. As soon as you talk about a network as a living system or a living organ then it is doing something.
A Yes, but a network can exist Without influencing outputs in the sense that you just used the term. The example of the « invisible college" is one which I keep in mind. A group of people who share papers, ideas, and so forth could conceivably do that without the concept of input as we know it. It is the people individually who are perhaps getting the input. They are generating the information which flows through the network. It doesn't ever have to go outside. Now you could call it a closed system I suppose, but that diverts the nature of the discussion.
H We shouldn't quibble about terminology. What we should try to get at IS the concepts involved. A network is something that consists of nodes and arcs. Now just for terminology's sake, let's say that a system consists of components and links between them. Now there is no reason that you cannot view a system as a network. You can say this system is a network of components. you can say that if you want to. My only argument is that in order to make the concept of system meaningful. each component must itself have dynamic properties. In other words, it does not get you anywhere if you create a large system that is just static.
X A network can just sit there. You can conceptualize a network without attributing any particular movement or change to its components.
Y Is not there a question of control involved here when you were talking about the system earlier on ? There was a notion that the component of the system had things coming into and out of it, it was transforming them?
A Would you then say that the nodes of a network do not change? The only thing that can happen is that the network grows by producing more nodes or less nodes, but the nodes themselves do not do anything. The most consistent thing is the transforming.
H Strictly speaking yes. There's no reason you can't have a network In which you take the nodes and you put something in each of the nodes, something that does something but that does not change the structure of the networks. You see, the structure of the network is intrinsic to the network itself, whether those nodes do something or not. A good example is the kinship network, because you can define a kinship network without worrying about what the individuals are doing. In other words you just draw a family tree and there it is.
Now, you can also consider each of these nodes as a functioning individual and that individual is some kind of system or organism, and you can also have these people relating to each other by human behavioural interactions.
A What I hear you saying is that a network is simply a representation of a system. I mean I know that it's not and you probably won't agree with that, but that is what I am deriving from what you are saying.
H I am saying a network can be viewed as a system but that the network has a fundamental structure which is something peculiar to the network whereas the structure that we attribute to systems is an abstraction that we've created. It is a convenient type of modeling device that we've created. in order to separate the functions of a certain whole or organism, or portion of society, or whatever The boxes in a system representation are abstractions we've created. We decided that there are certain components or certain functions which are being performed. and it's useful to conceptualize them this way. A network is a more fundamental structure which we are somehow perceiving in the thing that we are trying to model. We take that as given before we even talk about change or dynamism. It's a fairly simple distinction
X In both cases you have an abstraction. The same natural system can be conceptualized as various networks or as various systems It seems that you think that the process character in the system case is not only in the flows between the boxes, but also in the boxes. Whereas in the network case, It could only be in the connections, the nodes themselves are not doing anything.
H I am saying that these nodes themselves could be systems, could have system properties attributed to them. In fact they can be subdivided.
X So all you say is that they are not analysed further at the given modeling stage; that would also hold for each about system box.
H It's not necessary to analyse them in order to create this model.
X Tell us about topology. I think maybe we are worn out.
Y I don't agree With you that this is a waste of time. I am beginning to think that it's really quite important, because after all we structure thought through language and sometimes we talk about systems and we aren't sure what it is exactly we are saying. We talk about the telephone «system» and yet we talk about the telephone « network » and we mean something different by those two concepts, but in spite of all this time we've spent if I had to write an examination on the difference between the two. I would find it still difficult.
Z Could we address ourselves to the telephone system versus the telephone network?
X Presumably one can prove something with this network theory and presumably the kind of things you can prove with network theory are different from the kind of things you can prove With certain kinds at system theory.
Y If you want to talk about people you could talk about the social system versus the social network and you'd be talking about quite different things.
Z We might be talking about the same thing in different ways.
X Could we also look at this. The main reason why computer people find that systems work better is that they represent the flow of control.
R What about their use of data « network » versus computer « systems. »?
H Let's talk about telephones. Now what is a telephone « network» ? Very simply a telephone network is a bunch of little phones that are connected together. There is a phone in every house and there are lines linking them and some of the links go through central clearing houses and so on, and then they go back down to other phones, a telephone. system» is something much more complicated.
E So the physical network is there. It mayor may not be used. It may only be used from time to time. Just like a ditch in the ground, it's there. Sometimes the rain gets in and water runs in it. You can break one of those links - you can move a node, etc.
H The telephone system is something much more complicated. In fact wouldn't presume to define that in any particular way because any individual In Bell Telephone might define that differently, depending on his function one way to look at it would be in terms of the telephone information system, in other words how is information transmitted, how is it routed.
It's all done by electronic hardware Now there is another level you can look at the system, that's the human decision level and that goes a little further how are rates set? How are lines administered ? How is policy made in towns of how phones are maintained, how they am installed? How is the whole organization created which will support this? The former can be automated but not the latter. The key distinction seems to be in effect the network gives you the set of constraints.
Z When you didn't have automatic exchanges you worked through operators and there were certain limited possibilities - the rate of flow of messages, the manner of using them and all the rest of it. The system is conditioned by the constraints that the network describes. The minute you speak in terms of greater flow and the way it's being used, you are talking about a system. In this sense, most of the policy decisions in the telephone system do not take into account all the actual nodes of the network. What they do is they take a lump sum, the averages, forecasts. They think of them in a conglomerate sense. They are forced to because they can't possibly look at everyone.
A It would be really helpful if you took the same distinctions and transferred It out of the telephone network or system to say a political network or system. Let's try that.
H You can think of a social network as a network of individuals, do you want to look at it at that level. It's up to us to choose a level: people, organizations, states, etc. So here's individual people, p1, p2. etc. Now in this case it's a bit more complicated because you don't have the fixed links. You might want to consider it as a dynamic network in which different people are associating with each other at different times. That gets kind of complicated. The minute you itemize it you are talking about a system. If you talk about a network that changes at each instant then you are right, it becomes very difficult to deal with it, certainly on a theoretical basis.
A What I thought you were leading toward was a definition that would say that one could define those links between PI and p" as not being active all the lime. They don't require any physical connection. They really are are like the ditches before the rainfall. They are predispositions, they are communicational predispositions so to speak. But they have characteristics.
E You know what bothers me every time we discuss the nodes we get into trouble because there is an unknown quantity like what happens when we don't get free passage of information, or we get distortion. So because you have a change of a message going through, then you say you no longer look at that as a network. You start worrying about maybe that's a system.
H I don't see why that should bother you because in effect the node is a link. It has the same kind of problem. You've got to look at it in terms of the physical constraints or institutional constraints. Within that you then determine how it operates: which is the system aspect. If we want to make a distinction rather than engage in semantic exercise, we might just say that for our own purpose, the minute we begin to talk about the «thing» that has variations in flow. or patterns, we are talking about "systems" but when we refer to "networks" we really have to say what the constraints are within which all that effort operates, whether physical or biological (at any point in time), because moving from the operator to an automated system means that we have changed the technical constraint.
X We talked about time variant networks and certainly in electrical network theory all networks have time variance, they keep changing.
Y So that the networks themselves would change but not what they are doing. What they are doing does not take place in time If they do anything that takes place in time they become systems.
H I think we are quibbling over terminology here. You could call it a system but it still retains all the basic properties of the network, so it's a time variant network.
A We are only quibbling over terminology if in fact we confine this to the mathematical description. As soon as we try to apply that to social systems, we are not quibbling about technical terminology anymore. We have to know the difference.
H I think if you look at it in the application context of a particular problem then we'll find that we won't be worrying about the words we are using. It'll become clear as to what kind of model suits It best and whether you choose to call it a network or a system-type network
X It might be easy for us to represent something which doesn't change with time as a network, but we may be obliged to use such system representation when we are talking about something which does change.
Y It's not only time, It·S also complexity that's important.
B You made a distinction between a pattern and a process and it seemed to me that if your distinction is that a network is a pattern and a system is a process then I would disagree with that in the sense that a pattern is also a process only it's of a different duration.
Z A network can be described apart from its function is a sentence on which we ought to agree or not. There's no such thing as a system apart from its function.
X What about the solar system?
Y The solar system functions but not in a purposeful manner. Well, purposeful in terms of man's purposes. But there are certain systems that we can describe and that we cannot control, and this is where the question of control comes in. They do vary with time, but we don't necessarily have any says as to the process, as to how they function. However we can describe them and use a systems-type of description. I guess. we could differentiate in the sense that a network can be static whereas a system is usually or invariably dynamic.
H Consider a chart that describes a dynamic system. The chart itself may not change but what is going on. what is represented, is a changing system. So we have a diagram that describes a static structure initially. Now once we've defined the terms of that structure, we can then presuppose certain kinds of changes. even evolution of the structure itself.
X Does it become a system?
H It doesn't have to become a system unless you choose to view it in this way, and then you would no longer be designating individual units as the components of your system. You would be designating some kind of processes or relationships.
A Can I ask you a question? I think we've sort of been haggling around this and I don't believe we really should. Is the way you're describing it a more or less accepted way of dealing with it among mathematicians.
H I'm giving you the « party line ».
A Now the question. We may be In a different party. We may have a different party line. It's important for us to know. Will the biologist accept this? The mathematician does. The mathematician accepts this, now will the biologist accept this?
C Well. the biologist can choose to use that.
A Will the economist accept this? He would.
Y The interesting thing it though that there is no party line. I did an overview on the use of systems analysis in the federal government for an applied systems analysis conference last year. 1 had a researcher going asking all the departments what do you dot that you call "systems analysis"? Everybody called what they were doing systems analysis because it's now a sexy thing; they want to be in on it. What we did was to differentiate in terms of purpose. namely"whether they identified the control variables, so that some people making a decision could find their activity of some use. This was the key in terms of what was usefully called systems analysis.
Z We are getting back to the question of control.
A This would be useful because some of us were worried about it, partly because of people constantly asking questions as to whether we are playing with words or whether we are discussing a real distinction between networks and systems? The answer surely has to be I don't know. Maybe no one can answer that question. But what we can usefully do is say that what we do mean by networks is this, and that we are convinced we're not playing with words.
B What I am interested in is the new sexy use of networks. It is now being used because it's sexy. Why do the people use it in this new way? I want to make a distinction between that use and the widespread use of systems.
Y For what it's worth I never heard, in all the discussions that I had interviewing all those people, any use of the word network. Not even once. They take it for granted that what in effect you are talking about is relating boxes, decision makers, or data, phenomena, and when you set up your system you've got to specify these relationships.
A On the question of control, I want to stick to that because of what the degree of control differentiates from a policy point. From what you've just said, would you accept the notion that when you aggregate from networks you tend to get systems?
H I would agree with that in general. How do we control networks? First of all, let's look at the ways we could do It. Let's not think of a static network but of a network in which there are interactions taking place, so that the individual components are occasionally interchanging information, or it could be goods or whatever. A network is really only a snapshot at one moment of time. You could have a picture which represents the interactions over a period of one month or you can have an instantaneous picture.
A Are we agreeing on that, are we buying that network is simply a synchronic slide?
H I think someone mentioned earlier that certain forces within the network may be active at certain times but that the network as a Whole can be considered as representing all the potential activity which is possible, all the potential relationships which are possible. Then at any given moment a particular snapshot will give you certain values of the parameters in the network. For example, the intensity of the activities of a certain person or node, or the intensity of the activity across certain arcs.
X We are all caught up with the idea of potential, that a network is a set of potentialities and that only when they are called into play are you going to get anything out of that photograph.
H A network can describe the total potential of a complex system and then at each instant of time we can describe the actual activity level of that system. Anyway, how do we control it? One way to control it is to change the relationships between certain components of the network, either by inducing some kind of closer association or by cutting off the association some way. or perhaps by changing the nature of these associations. Again, I am speaking very generally if you think of a social system, you could try to make transfer from this person to this person easier or more fluid by somehow changing the process by which they do interrelate. That might be by facilitating communication or just by having a T-group session and teaching these people how to communicate with one another. Now I'm not going to call that control. It's not a precise kind of control in the sense you are not manipulating. All you are doing is facilitating, or you are somehow altering the properties of the network so that it will achieve a function which is more in line with what you desire. It's a way of working towards a goal.
X The control is being done from the outside of the network is it? You describe an outside agent, whereas in the system case control is inside the system.
H Well. we can talk about system afterwards. I think that in both cases you can have control either within a system or without, but to me the idea of control implies someone external to the system who has some overall goals and is trying to affect them by playing with the system somehow.
B The one thing I like about human networks is that there are people interacting with one another toward a common end. I think one reason they've come into existence is because they are difficult to control. They are difficult either for a single person or for an outside group to control. In a human system, if you attempt to control one of the links, uncontrolled links will be found or created. It's adaptability or robustness. Networks, as opposed to hierarchies, have alternative paths so that nobody can seize them. If you seize them if you seize a network in one place, it gets away in another. People feel more secure operating in a network.
Z Is that a necessary Characteristic? And If so, isn't that redundancy?
X Are you talking about this tendency to create alternative paths?
A I am thinking of real networks as I know them and trying to abstract and see how they are represented. to see whether they can be represented in these terms.
Y Well. I think the subtlety of control has a lot to do with it, in other words, the behaviour of the human animal is such that, as we mentioned before, he can resent any obvious manipulation.
X But let's talk about trying to control the network as a whole. The network functions for a purpose. People join up to stop something, or plan to exchange information about ecology or physics or something else. Suppose you want to stop an exchange of information you can have a very difficult time because you are going to have to really end the whole network.
Z When you can anticipate all the alternative pathways then you are in a better position to control.
Y In economics if you are ever going to evaluate, you've got to exercise your imagination hut in order to make it a meaningful policy-relevant exercise, that imagination has to have a sense of network, that is a sense of the potential relationships as well as the actual relationships that might exist.
H Well. in any case if you are trying to control a network it will probably not help just to alter certain portions of it, you'd probably have to effect the change in all the links and all the means of relating the transfer functions that exist.
A I see that as a useful observation about networks.
B I'm not convinced that there are examples of networks that can be controlled. Let's take the telephone network versus the telephone system. You can talk about the Bell system which is something the Bell Company can control in the United States but they cannot control the telephone network and the network which links the Bell network (the U.S. part of the network) to the European part. The telephone network stretches inter-continentally and only part of that is controlled by the Bell system. So the network cannot be controlled, I'm looking for a better example.
A Would you agree that networks cannot be controlled?
H Well, I'll say that if you can isolate a network. if you can definitely put boundaries around it and say that's your domain of interest and that there is nothing outside the network that could affect it. then you can control it. But the trouble is anything that we try to model as a fine network always has little connections leading outside of that domain.
B A good example I think are the innumerable cases in which secret police " neutralized" political revolutionaries by putting them in jail, by suppressing some of their links, but effectively parts of the network probably continue to exist sometimes for 20-30 years They may be reduced to ineffectuality, although sometimes the network reconstitutes itself. But is it possible to control it? It would seem to me that for long periods of time you can, but maybe by not only suppressing certain links and isolating certain people in that network, but also by being aware of temporarily inactive elements in the network.
Y From what you were saying I would say that you could destroy it, using your example, by just eliminating all the nodes, in this case the people.
X Well. you may not be able to exercise complete control but it turns out in the case of Russia that for about 17 years Lenin'S principal aide, Roman Malinowsky, was a police agent.
Z He was controlling the network?
X He was certainly influencing it sufficiently for long periods of time to make certain things come out the way somebody outside the network wanted them to happen.
B I would accept that but I think there's a big difference between influence and control. In the end of course the network took over.
H What we mean by influence is not control. When I use the word control I meant influence on a day to day basis so that the progress, whatever the organism was trying to do, was shaped in a certain way. In the example of the network of people trying to stop nuclear plants, alternatives were always chosen which didn't threaten the nuclear plant. That was a good example where the network allowed itself to be controlled by an outside force.
B I can see that there is a distinction to be made here. The first part of that example described a network which did a variety of things. It had a variety of purposes. It kept generating new purposes. At that stage it was a network. When we get on to the second part, suddenly some of them had this new purpose, namely the anti-nuclear thing. When this purpose emerged when the plant was going to be installed, when it became a threat - that galvanized the network into a particular configuration.
X I understood that the issue created what could be termed a whole new network - they had people who never worked there before - new nodes.
B But it also had a special kind of structure which was explained to some extent. As far as I am concerned my interpretation would be that part of the network was galvanized into a new form which I would consider to be a system or organization. The way those links were drawn the information was passing out from the centre to the periphery It was admitted that they were organizational links.
Y And people from the periphery travelled to the centre to do their communication.
B Yes. but there wasn't horizontal communication between the different elements of the periphery which is part of the characteristic of the network as they originally defined it. So I think there are two levels. This is in effect the core of the distinction.
Z The issue changed the form and the network became controllable for a particular purpose.
A Could we just broaden the notion about slipping from the network to the system? Could we perhaps just hear about the controls in the systems in order to get this conceptually together?
H OK, I guess we've made it clear that if you try to go into a network and just alter it by just changing certain patterns you are likely to have a lot of consequences that you do not predict. Maybe continuous control is one way of avoiding that but that takes a lot of energy. You need a very complex control mechanism.
A Aren't we saying that any system or any network operates within an invisible web of constraints be it legal, social, cultural, etc. and there is a kind of space within which you operate which changes, of course from moment to moment, and once you've specified that network you are not going to be able to describe the system to control it.
H Consider this case let's assume that you have control. A welfare system sets up certain goals namely that whoever wants to see 3 doctor should be able to see a doctor. We have a constraint in the sense that there are just so many doctors. Then you find that time per patient goes from 30 to 15 minutes because of the way in which the system is being used. But we then set another constraint. namely no patient on the average will have less than 15 minutes. When we do that we then have to control people We then have to say within these goals, constraints, configurations we don't know what would happen if we put a dollar deter. ant fee. Now you then have a different network when you add that particular variable. The network is how people will act: patient to doctor, patient to hospital. But the rate at which the patients move is a system function.
B You're associating quantity with system, quality with network.
H Well, the only problem there is knowledge or information mushrooming to the extent that you can't control all the information that you have. To design an ideal system you want to use as little information as you can and control as much as you can, or know as much as you can about what is happening. The problem is selecting the right points at which to get you information.
X Maybe one could look at some of the situations where one would intentionally use network rather than systems analysis. I find systems "sexier". They seem to be able to do everything.
H They are much more general. A system description can be used for almost any kind of dynamic processing: social. biological. organizational systems are much mom flexible because the concepts are so broad. A network is a very specific type of structure and people often misuse the notion of a network where it just does not apply.
A But hang on a second, you are quite correct where you can define the boundaries. When you can't define the boundaries you have to go back to network.
H If you can't find boundaries, that may mean that you can't do anything at all. It doesn't mean the network will necessarily help you anymore with assessing the problem.
B One of the problems that I have come across is that in dealing with many different kinds of organizations you can define what's happening within each organization, but you run into big problems when you try to find out what happens between organizations.
H OK I can see what you are trying to say. In other words in a system you want to get a very detailed description of the behaviour and interaction of all the components which may not be possible, whereas you could create a network model which at least shows the existence of those components and the fact that they are related in some distant way whatever it is. In a system you need a detailed description of the behaviour or the performance of all the components as well as the interaction between those components. It may not be possible to obtain this kind of description. You see the boxes which we use are really short-hand notations for aggregations of people. Even in the case of an individual, we don't know enough about the human psyche. The person's mood may change from day to day, and so forth. You are dealing with generalities, the predictability of which is not too high.
C I think a very nice example of this IS a biological system. We were taught to model a cell as a system and what you get to realize is that a biological system is so complex that it's almost impossible to figure out how to control it, because you cannot really describe it. Even a simple cell, is incredibly complicated, let alone a conglomeration of cells. And yet we know that these constitute a system, and we know that it works. In fact it's a self-regulating system. There are all kinds of enzymes and various glands and organs manufacturing the proper amounts of chemicals to maintain this incredible balance throughout the body.
B Have you seen the metabolic pathways chart? It's a big chart about 150 X 100 cm. It represents all the processes within the cell and the links between the different enzymes. Now what intrigues me is that I don't think that could be defined as a system, but It can at least be represented as a network.
C This is essentially what we try to do, namely to represent the cell as a network, realizing that we can never hope for an accurate systems description of it.
B What you are saying there sounded really good but then as I began to think about it. What you are saying is that a network is an ill-defined system. It's a system in which we simply can't specify the interaction between the components or the nodes.
H No, what I was saying was that if you have 8 system which cannot be well-defined, you can at least give an ill-defined network representation of that system. It's not to say that all network representation must be ill-defined. In physical sciences you can define it more precisely.
Z All systems are natural but some of them are more natural than others in the sense that self-regulation appears in forests, in cells, in something which pertains to the biosphere itself but it doesn't seem to appear in what you were saying about a telephone network. And why? Because telephone networks are a small fragment of a big complicated system. We sort of look at the local complexity and so I think ... (end of tape)