Outline: Date: Mon, 26 Feb 1996 17:24:01 -0400 (EDT) From: DON MIKULECKY Subject: A good set of question with the best answers I can provide...anyone????? Date: Mon, 26 Feb 1996 18:48:30 -0400 (EDT) From: DON MIKULECKY Subject: here's a rough rough draft ...see if you wish to comment or join in Date: Tue, 27 Feb 1996 10:37:01 -0400 (EDT) From: DON MIKULECKY Subject: some questions from seth {My reply in these brackets....DCM} From: GEMS::SBROBERTS 26-FEB-1996 22:20:45.91 Subj: question about relational models From: GEMS::JPRIDEAUX "Jeff Prideaux" 27-FEB-1996 13:55:54.13 Subj: thoughts on constructivism From: GEMS::JPRIDEAUX "Jeff Prideaux" 27-FEB-1996 15:59:47.67 Subj: RE: thoughts on constructivism Date: Tue, 27 Feb 1996 22:23:34 -0500 (EST) From: Jerry_LR_Chandler Subject: Re: Jeff is cooking. I like when he speaks with his own words!!! From: GEMS::WILSONJD "John Wilson - Radiation Biology" 28-FEB-1996 01:03:45.19 Subj: morowitz, hal From: GEMS::WILSONJD "John Wilson - Radiation Biology" 28-FEB-1996 01:09:38.94 Subj: hal II From: GEMS::SBROBERTS 28-FEB-1996 08:12:16.79 Subj: RE: Some thoughts from Jerry Chandler From: GEMS::JPRIDEAUX "Jeff Prideaux" 28-FEB-1996 08:44:34.68 Subj: C.A.S.A. and constructivism Date: Wed, 28 Feb 1996 13:20:50 -0500 (EST) From: Jerry_LR_Chandler Subject: CASA - and contructivism In-reply-to: DON MIKULECKY Date: Wed, 28 Feb 1996 18:26:02 -0400 (EDT) From: DON MIKULECKY Date: Thu, 29 Feb 1996 09:04 -0500 (EST) From: CARROLLT@PCMAIL.CBIL.VCU.EDU (CARROLL, TIMOTHY J.) Subject: Re: welcome! Date: Thu, 29 Feb 1996 09:47:19 -0400 (EDT) From: DON MIKULECKY Subject: where are we headed? ____________________________________________________________________________________________________________________ Date: Mon, 26 Feb 1996 17:24:01 -0400 (EDT) From: DON MIKULECKY Subject: A good set of question with the best answers I can provide...anyone????? Dear Jerry: {I'll have my comments in these brackets} Dear Don: In light of Jeff's comments, which I only partially understood, I thought this comment on reductionism was worth further analysis by the group. In addition, I remain confused by exactly what Rosen was attempting to prove with regard to "efficient causality". Perhaps you can give me your meaning. In particular, what do you see as the theoretical consequences of Rosen's conclusion that life is not a mechanism? And what are the practical implications of the theoretical conclusions? {Well, I hope you realize that this is the essence of his book: "Life Itself". I will try to give a brief answer. The difference between an organism and a machine is very practical. The difference lies in the issue of being closed under effecient cause, but this has many ramifications, all of which are practical. Being closed under efficient cause is a property that guarantees that no component of a system is unentailed (without a way of being realized or made, i. e. caused). Here the word "component" is crucial. Complex systems are neither "top down" nor "bottoms up" contrary to the popular notions coming out of the Santa Fe Institute group. The issue is that functional entities called components do not necessarily map nicely into the parts one can physically isolate. Therefore they are irreducable and they can not necessarily be retrieved by merely reassembling the parts. Machines have no distinctions between parts and components. They have what is known as a "largest model (usually dynamics) from which all other models can be derived. The causal relations in a machine are such that as you put it back togeher from its parts, the last stage is unentailed. Thus machines need a creator as Descartes was willing to conceed. Organisms do not have this property and the most important consequence of this is that no attempt to reconstruct a picture of an what an organism is from machine like models can succeed!} If you can find time, I would also appreciate getting copies of your technical papers in network thermo - or at least some references to them. {I'll send what few I have left, but the best source now is my book: "Applications of Network Thermodynamics to Problems in Biomedical Engineering" NYU Press, 1993. ISBN 0-8147-5490-2} In the case given below, as well as 'reductionism" in general, it is difficult for me to decide where the rigor of science ends and the beginning of philosophical speculation begins. History suggests that a crisp representation of a system and well defined terminology are critical to good science. Living systems should also be amendable to this approach, I believe. {That is precisely why I spend so much time preaching Rosen. He is so precise and so careful and uses the very best tools available to "science" to show that what we used to call science is inadequate and falls short of its own goals and standards. There is no line of demarcation between the rigor of science and philosophical analysis of science's limitations if careful procedures are followed. Scientific speculation, even when dressed as rigor, is as weak as any other form}. In light of the review comments you posted, I look forward to learning more about your conclusions from a network perspective. I have drafted a paper on thermodynamic system categorization ( draft #19 has existed for over a year; draft 20 has recently been initiated). Thus, I am most curious as to your views. {I'd like to see that! I wrote a bit today on these issues and will circulate what I have as soon as I am done with this draft}. Sincerely Jerry ---------- Forwarded message ---------- Date: Mon, 26 Feb 1996 13:45:37 -0500 From: Kathryn Blackmond Laskey To: krasnow@GMU.EDU Subject: reductionism Dear Krasnow list, I thought it would be interesting to raise the following issue for discussion: I'll start with a quote from the call for papers sent to this list today: >* Complex systems may resist reductionist analyses. In other words, it >may not be possible to describe some systems simply by holding some of their >subsystems constant in order to study other subsystems. {No! Complex systems do resist reductionist analysis. This has been rigorously proven by Rosen and others.} For a long time there has been a great deal of simplistic and not very productive bickering between "reductionists" and "holists." Sound bites such as "Reductionism -- you can't do science without it!" versus "Reductionism means killing it to study it!" characterize this debate. But we seem to be evolving toward a new conception of what reductionism means. This new viewpoint says that reductionism *is* necessary to doing science -- if by reductionism you mean trying to understand a complex system in terms of interactions of simpler subsystems from which it arises. But except in the very simplest of cases, interactions between subsystems prevent you from "holding some subsystems constant in order to study other subsystems." {The rub is in how one defines "subsytem" If you simply mean isolateable parts this is wrong. If you mean functional components, yes some reduction may be possible. In the Benard system, there is clearly NO reduction possible!} We now see an explosion of the following kind of research: You build a computer simulation which models a complex system as a bunch of interacting simpler subsystems, and then you try to abstract from the results important properties that apply to the real world system under study. I claim that this activity is reductionism, although a more sophisticated sort than the "holding other systems constant" variety. After all, aren't we modeling the complex system in terms of simpler subsystems? {By definition, computers CANNOT model complex systems, since complex systems contain non-computable aspects. What computers can do is mimic the behavior of a complex system in a way which is definitionally unfaithful to the causal relations in the real system. This is the fallacy of "Artificial Life" studies!} But many important questions remain about how to do this sort of activity scientifically. When is it legitimate to say we have made a scientific advance from such a simulation exercise? Is one of these simulations a theory? An instantiation of a theory? What does it mean to instantiate a theory as a simulation? How do we know whether such a simulation-theory is falsifiable? {since simulation is always a form of mimicry, only the validity of the mime can be established in this way. Simulation can not validate theories, only a commutative modeling relation can.} Any statistician, or anyone who plays with complex simulations for a living, knows that if you put enough "knobs" into a simuation model you can get it to do just about anything. In many of these models, it is *very* difficult to know what are the "knobs" on which the results depend. If we are to believe that some abstraction from simulation results applies to some real-world phenomenon, we have to be able to argue that the result does not depend on unimportant "knobs" in the simulation, but is a general structural property following from the theory underlying the simulation. This is a great intuition, but how can it be made rigorous? {only by moving from simulation to careful model building} Is anyone interested in getting a discussion going on this issue? {we already have one....join us!} Kathy Laskey Best wishes, Don Mikulecky ____________________________________________________________________________________________________________________ Date: Mon, 26 Feb 1996 18:48:30 -0400 (EDT) From: DON MIKULECKY Subject: here's a rough rough draft ...see if you wish to comment or join in This is a very rough draft of some ideas I have been massaging for quite some time. I wrote them down with the hope that Eric Schneider might use them or else reject them as a first step in our putting together what we are about to write. I circulate it with the hope that at least some of you will read it and comment or even join in on the venture. On Organization, Thermodynamics, and Complexity. D. C. Mikulecky The concept of organization is central in so many discussions of complex systems, yet it defies any commonly accepted definition. In particular, the notion of "self-organisation" has become a buzz-word. This status carries with it an ever present danger, namely that the word itself will loose meaning through overuse and fashion. It is our purpose here to examine the concept of organization afresh using tools which are readily available and which help add a precision to our thinking. One of these tools is network thermodynamics [Mikulecky, 1993], which, as we will see, focuses on organization in systems in way which allows us to be more precise. First, it is useful to review some history from an epistemological perspective. This will focus our thoughts and define the problem in as clear a way as we are able. Two branches of physics were brought together in a form of "epistemological reductionism" [Peacocke, 1985] in order to show that thermodynamics is "merely" a way of dealing with cetain select averages over the multitude of coordinates in the dynamics of a large collection of particles. From this was born the field of Statistical Thermodynamics. Its results were very satisfying to anyone wishing to be reassured that powerful ideas like the three laws of thermodynamics had their roots in Newtonian particle dynamics and nothing more. Even the nasty epistemological problem of the irreversibility of observable, macroscopic events coming from equations of motion which were time reversible seemed to have its resolution in the nature of how the averaging was done. Rosen tells us in many places that physics is special and that biology is replete with situations which are not amenable to being fit into the special categories which physics handles so well. This seems to be a message which has bearing on the relevance of thermodynamics to biology. Some authors [Kaufmann, .....] seem to wonder whether or not thermodynamics even applies to living systems at all, given some of the popular interpretations of the entropy and the second law. Some even question whether the second law should be considered a law of all nature or whether it needs to be limited to the non-living world. The proof of the second law due to Caratheodry [see appendix in Mikulecky, 1993] is one of the most elegant proofs in all of science. It rests on reasoning which is purely relational using tools from the branch of mathematics known as topology. It needs one experiment in the real world to make it a universal truth! That experiment can have any number of realizations, but one which always seems appropriate is the Joule experiment to find the mechanical equivalent of heat. In this experiment, a flask is fitted with an insulating jacket to prevent heat exchange with its environment. A paddle/crank apparatus which is designed to be as close to frictionless as possible is attached to flask in such a manner as to allow the paddle wheel to stir water inside the flask as the crank is turned from without. The requirement for Caratheodry's proof is that there be some state which is innaccessible from a second state along an adiabatic path. That is a path which changes the state of the system in a matter consistent with the constraint that no heat travel across the walls of the system. In the Joule apparatus this condition is met simply and elegantly. The "operator" turns the crank doing mechanical work. The effect INSIDE the system is an increase in temperature due to an addition of heat energy. No heat energy was transferred, the device CONVERTED mechanical work into heat inside the system. From a particle motion perspective, this was the "randomization" of the motion of the water molecules as they left the faces of the paddle wheels. The distance a water molecule can move in liquid water before colliding is less than the diameter of a water molecule. This is a demonstration of the notion of IRREVERSIBILITY. This randomization is not reversible. The "conversion" of directed mechanical particle motion into the random motion characteristic of "thermal" energy is not capable of being reversed. We might say that an ORDERED motion becomes disordered. This is a good model for the nature of frictional heating in general. We call this DISSIPATION. Thus due to the irreversibility of this change in the nature of the particle motion, WORK becomes HEAT and the initial state is TOTALLY INNACCESSIBLE from the final state as long as the insulating jacket remains intact. Caratheodry's elegant topological proof has its one necessary realization in nature and the rest is mathematics! This demonstration of how a universal law can arise in so simple and so elegant a way has a potentially esthetic beauty to it. In spite of this, it probably is one of the most underappreciated and misunderstood concepts in science. The number of deep ideas associated with this demonstration is remarkable! The mathematics goes on to demostrate that because the inaccesability condition is met, there is an "integrating factor" for the heat energy function's differential form and that this allows the definition of a new state variable which we know as the "entropy". With just a little more effort, it is easy to show that the integrating factor is the absolute temperature. By employing statistical reasoning via the statistical thermodynamic considerations mentioned above, the link between the randomization of the particle motion, the irreversibility of the state change, and the INCREASE in the entropy of the system as equilibrium is achieved after the perturbation are all now identified as part of a whole idea in a consistant way. It is only a minor excursion to incorporate Boltzman's statistical interpretation of the entropy as a function of the number of ways a given state can be realized to complete the line of thinking. The result is the recognition that because the distribution of the states accessible to a system is sharp, the equilibrium state has the most ways of being realized by far, and therefore is the most probable and the one with highest entropy. Thus another principle is born, namely the fact that equilibrium states are simultaneously those with maximum disorder and maximum entropy. Hence entropy becomes a measure of disorder. Order vs. Organization So far the domain of the development of concepts has been very carefully restricted to extremely SIMPLE systems [Callen, 1960]. Here we use the word "simple" in the sense intended by Callen, namely systems of one or at most a few chemical species, with homogeneous composition and little or no structure. The next level of sophistication comes from putting these simple systems in contact with each other using idealized walls or interfaces which carefully restrict the kind of transfers possible across them. This provides the best of all possible worlds. Statistical thermodynamics allows the epistemological reduction of all thermodynamic properties to Newtonian or even quantum descriptions of particle motion. Then the real world is approached by introducing ORGANIZATION into the picture once this has been accomplished. Notice that the order/disorder dichotomy has to do with the randomization of particle trajectories in simple systems, while the introduction of organization is a matter involving the juxtaposition of simple systems. In principle, this build up of more and more organization can be carried on indefinitely. In practice, things get complicated fairly early in the process and systems are generally fragmented to a small enough number of simple systems as to remain tractable. One fundamental error which recurrs from time to time is that the orgainization of a highly structured system is simply ignored and the statistical thermodynamic epistemological reduction is applied to it as if it were a simple system with no structure. This enterprise is one doomed to introduce confusion and misconceptions, but that has not limitited its reoccurance it seems. One semantic difficulty which can arise in this context is the confusion of "order" with "organization". It was not until the work of Peusner [1985] that a method for the application of thermodynamic reasoning to more highly organized or structured systems from the equilibrium thermodynamic point of view was discovered. This was an offshoot of his successful application of network theory to non-equilibrium systems [Peusner, 1970, 1986a] in the form of NETWORK THERMODYNAMICS. What was most important about this new approach to organization in thermodynamic systems was the demonstration of the applicability of simple network concepts so succesful already in electronics to systems of all kinds. This had already been recognized in some areas of engineering in the form of "physical systems theory" [Blackwell, 1968; Athans, Dertouzous, Spann, and Mason, 1974; MacFarlane, 1970;Roe, 1966] but network thermodynamics went a considerable distance further. Peusner recognized that the reductionist reasoning which led to the development of statistical thermodynamics had also permeated Onsager's thinking when he extended thermodynamics into the "near equilibrium" non-equilibrium domain by considering fluctuations around equilibrium [Onsager, 1931a&b]. It is not surprising that Onsager reached out to the epistemological reductionist approach of statistical thermodynamics when he sought a proof of the reciprocity between effects of one part of a system on another in an interacting system. Others have commented on the fact that this appears to be just a manifestation of Newton's third law of motion, so why not look for its origins in particle dynamics? A careful reexamination of Onsager's efforts using the insights gained through the advances made using network thermodynamics set the stage for a totally different way of dealing with thermodynamics as an approach to systems. First of all, the proof Onsager sought is relatively simply obtained using a theorem from electrical network theory called Tellegen's Theorem [Mikulecky, 1993]. All that is required is that the network in question obey a generalization of Kirchhoff's laws. These are simply conservation and closure laws which are obeyed by ALL physical networks! More important the nature of this proof is topological in nature. It is based on relational information, conservation laws and NOT on particle dynamics in any way! It is worthwhile digressing into the social aspects of science for a moment. Peusner sent the proof of Onsager's reciprocity to Phys. Revs. where Onsager had originally published. His paper was rejected on the grounds that it is IMPOSSIBLE to prove the reciprocity relation with statistical physics, in other words, without particle dynamics. When asked what was actually wrong with HIS proof, the editors replied that his paper was too long to read! As a result the proof was published in the Journal of Theoretical Biology [Peusner, 1986b]. Ironically, given the relationship of physics to biology (the special to the general), this may have been the best outcome. This digression is not irrelevant. The issue it focuses our attention on is the fact that the reliance on partical dynamics to explain EVERYTHING had become a BELIEF, not a fundamental aspect of science when practiced properly. This should be a clue to the theme which will be developed here. At the level of the actual transfer of heat from one simple system to another, the particle dynamics model seems appropriate and the associated notions of entropy, order and equilibrium seem adequate. As soon as ORGANIZATION enters the picture, the situation changes drastically. Now other, more relational ideas are the fundamental ones. Just as in Caratheodry's proof of the second law of thermodynamics. This fundamental, universal law has its origins in relational, topological ideas, not particle dynamics. It only begins at this point. The next issue which Peusner dealt with using the network thermodynamic approach is an issue which plgued thermodynamics for its entire history. As late as 1973, Callen, Tisza and others were meeting to discuss the "Foundations of Continuum Thermodynamics" at an International Symposium held at Bussaco, Portugal [Domingos, Nina and Whitelaw, 1973]. They lamented the failure of ever finding a metric structue to support the mathematical development of thermodynamics. In fact, there IS such a structure and it is inherent in the Network Thermodynamic representation of coupled, non-equilibrium systems. In fact, the choice of network reresentation made by Peusner gives the cannonical mathematical representation automatically! [Peusner, 1983]. Once again, the dicovery of the much sought after property came from topological, relational reasoning rather than particle dynamics! Once again, this is a beginning, not and end point. The structure of network thermodynamics is one which has taken us well beyond the limited story that can be told in terms of particle dynamics. It leaves on wondering whether or not particle dynamics is necessary at all. Finally, network thermodynamics is a sure fire method for generating a dynamics of its own. It is not a particle dynamics, but a systems dynamics. In linear systems, the result of a network thermodynamic formulation is a set of equation known as "state vector equations" which are well known to engineers. They are different in one significant way. The coefficients in the network thermodynamic formulation contain all the organizational information explicitly stored in combinations of incidence matrices. Once again an organizational feature of the system is what is contributed by network thermodynamics. This carries naturally over to the formation of nonlinear systems since the nonlinearities are ALL in the constitutive relations for the network components and not in the organization. Organization is capable of introducing nonlinear behavior in systems which are composites of linear compenents, when certain parameter dependencies are introduced [Mikulecky, 1988]. This has begun to sound like a pitch for network thermodynamics, but that is not the case. In fact, as in any approach, we learn as much from its limits as we do from its capabilities. Once we stated that the contribution of network thermodynamics to the formulation of systems description was a new, more information about organization rich way to formulate a familiar dynamics, we we had stated the limits to its usefulness to those aspects of a system we might call "complex". Why is this so? The answer to that question is best approached slowly and carefully. First of all, most of the popular notions of complexity attempt to identify it with lists of properties such as "emergence" and "self-organization". The problem here is that a kind of emergence can be demonstrated using network thermodynamic models of composite membranes. This goes back to the work of Kedem and Katchalsky [1963 a,b,&c]. Emergence is a concept which escapes definition in a clean way because it has so strong a subjective component. The subjectivity is in the connotation of "surprise". The surprise seems to be more due to having used a limited paradigm which generated false expectations rather than the discovery of a new property. We have known about boiling water for a long long time. How do any of its properties constitute "emergence"? Had we realized the limits of the Newtonian/Cartesian, reductionist/mechanistic paradigm earlier, there would have been far fewer surprises! Must a system self-organize to be complex? It would be tempting to say yes, but then what system does not have some element of self organization if we look hard enough? Again this concept is an "in the eye of the beholder" concept as much as emergence. We are forced back to Rosen's notion of complexity and are forced to ask what connection, if any, it has to thermodynamics. Here we seem to be entering virgin territory, in spite of the literature which would pretend to address this issue. Is it possible to shake free from the reductionist roots of thermodynamic thought and take a fresh view of nature from a perspective which retains those aspects of thermodynamic reasoning which are not simply checks and balances to make sure we don't stray from a grounding of all we do in particle dynamics? I think it is. We have already demonstrated that there are important concepts and theorems which are totally independent of partical dynamics. There are indeed more than those which we have mentioned here. What is missing from the perspective of network thermodynamics is the ability to model self-organizing systems. The network thermodynamic approach lacks any provision for context dependence. Networks have organization. At least one aspect of that organization is capable of being "encoded" in linear graphs and their incidence matrices. Clearly, this aspect of the formulation is totally independent of constitutive force-flow relationships. But this is not enough to model a Benard system. One can attempt "to cheat" and lay out a network topology which is ready to accomodate the outcome of the "mysterious" transition to convection cells, but this is indeed cooking the solution into the model at the start. The answer must therefore lie in another technique. But which technique. Certainly not one that derives from particle dynamics. Certainly not one which arises from a description of simple homogeneous systems. What else do we have? The Benard system has as its parts water molecules. That is all! What propertiy of water molecules is responsible for the transition to hexagonal convection cells? Be careful here since we can replace water by an entirely different fluid and still get essentially the same effect! Certainly a system with flowing fluid in it is involved in some aspects of particle dynamics as manifest in fluid dynamics. Certainly this system has some well defined thermodynamic properties as was nicely demonstrated by Schneider and Kay. But what is the reason for the formation of the cells? The system is anticipating its own destruction it seems. The system is obeying the laws of thermodynamics in a way which promotes organization while "trying" to fullfill the need to find more disorder due to the thermal gradient across it. Dissipation is increasing through the onset of self-organization in order to prevent further build up of a gradient across the system. [Schneider and Kay]. There is only one way to speak of this in terms of causality. FINAL CAUSE is at work here. This is the real complexity of this system. It is not the self-organization so much as the reason for it. The organization is doing something in the form of a new dynamic, still totally expressable in the old way, but existing for a reason outside of any dynamics we know. Now we see still a new dimension to our system.. It is not mechanism. It is not thermodynamics as we knew it, even as it was extended by network thermodynamics. If anything has emerged it is the need for us to change our way of looking at systems. The system opens the door for a more descriptive, less classical aspect to system definition and description. And it seems that no ammount of further calculation or model building in the usual mechanistic sense can remedy the situation! This system is not even close to being living. What does that tell us about the prospect of unraveling the secrets of living systems by the usual means? It seems that a new and radically different approach is needed. That approach seems to not only call for new ways of looking, but a discarding or at least an acknowledgement of the limits of our old approaches. This, I think, is Rosen's message to us. Complexity means that no number of traditional mechanistic descriptions can substitute for the simple recognition that this system, apparantly simple in Callen's sense, is one which contains a non-comutable component. One which we can speak of in terms of final cause. One which is independent of its constituent parts particulars. But finally, one which IS consistent with the basic laws of thermodynamics. Is there any more to be said? ____________________________________________________________________________________________________________________ Date: Tue, 27 Feb 1996 10:37:01 -0400 (EDT) From: DON MIKULECKY Subject: some questions from seth {My reply in these brackets....DCM} From: GEMS::SBROBERTS 26-FEB-1996 22:20:45.91 Subj: question about relational models Dr. Mikulecky-- am looking forward to reading your article, as soon as i can get myself from out under this respiratory test on friday....meantime, i have a question for you and the rest of the group.... as a follow up to something i mentioned in an earlier message, i still believe that in a sense, Rosen's relational diagrams are formalisms....the "rules" i was thinking of when i suggested that these models consist of symbols with rules for manipulating them, consist mainly of logical rules....so that you can't, say, have an arrow start in one place and go off in a hundred different directions...i understand what you said about still needing the modeller with these diagrams, but i think this is also true for other formalisms as well (the modeller has to create the mathematical relationship, say)....i think these models represent the kind of reduction that may be admissable for living systems...and in that sense, they are akin to other models of living systems....anyway, all this is superfluous... {Category theory IS a formalism. The diagrams you are refering to are what he calls "relational diagrams" and use category theory to represent relationships. These relationships are meant to be functional relationships between COMPONENTS in living systems. He also makes relational diagrams for machines. The machinesare diffferent. I see no way of algorithmicly constructing these diagrams. What rules, etc. are you refering to? The formalism of category theory has some, but there is an encoding into and a decoding from that formalism. As I see them, the relational diagrams are commuting modeling relations and, therefore, valid models.} my real question goes back to something i asked you and jeff a long time ago....that is, what is the relationship of Rosen's relational diagrams to the modelling relation?....is a relational diagram something that sits on the right of the modelling relation?....why or why not....the answer to this question would be very instructive for me..... thanks, seth {I think I answered that above...let me know if I was unclear} DCM ____________________________________________________________________________________________________________________ From: GEMS::JPRIDEAUX "Jeff Prideaux" 27-FEB-1996 13:55:54.13 Subj: thoughts on constructivism The following summerizes some of my thoughts on constructivism. I just wanted to get your reaction to this (after you have a chance to read it) No need to post this version. ------------------------------------------------------------ Constructivism What constructivisms isn't: Constructivism isn't merely "constructing" (or putting together) systems from the parts that had previously been identified through analysis. The paradigm of constructivism isn't "take a system, reduce it to its parts (reductionism phase), and then re-assemble (construct) those identified parts according to the 'wiring schematic' of the system". To understand constructivism, it is neccessary to understand the distinction between syntax and semantics. A modeler (who computational simulates a real-world process) goes through the following steps: 1. Determine what variables are important and write the equations of motion for these variables. This step involves semantic information in the mind of the modeler. This will be called the encoding step. 2. Integrate (or use numerical methods) to show how these variables change their values over time. This step is purely syntactic. It involves no semantics as to what the variables mean. No interpretation of what the variables represent is done in this step. This will be called the formal inference step. 3. The modeler now interprets what the results mean. This step involves semantics and relates the "results" back to the real world. This will be called the decoding step. Steps 1 and 3 involve semantics. Step 2 is purely syntactic. All three of these steps are important in forming a complete model of a real-world natural system. The model cannot just be identified with step 2. Since the formal inference step (step 2) cannot utilize any semantic information, the human modeler must intervene to augment the formalism any time a semantic change needs to occur. For example, consider that a modeler came up with a nice set of differential equations that perfectly described existing data about a real-world process. Upon integrating these differential equations, he/she found that they accurately predicted the future for a while. Suppose, though, that at some point in time, the real-world system acquired a new component that could not be interpreted by the differential equations (remember the differential equations [step 2] cannot know semantic information). At this point, the differential equations would be inadequate to predict the future unless the modeler intervened and adjusted the equations. The modeler could now say "well, if I had used this new set of differential equations from the start, then the model would had worked all along." The problem is that the real world system could experience another change at any time in the future and the modeler would be in exactly the same situation of having to adjust his/her differential equations. It is completely irrelevant how many times the modeler successfully adjusted the differential equations in the past. As long as real-time progresses, there is always the possibility that future changes to the differential equations will be necessary. Differential equations, in practice, are successfully used to describe only a very special class of systems---those systems that don't experience the appearance (and disappearance) of new observables that would then necessitate the encoding/decoding semantic steps in a model. In biological systems, new observables (for instance biochemical-molecules-aggregates) are continuously coming into and out of existence. Each time such a new functional aggregate appeares (or leaves) the scene, the event would have to be semantically interpreted (this cannot be done in step 2). In non-constructivist dynamics, it is the values of the variables that change. The variables themselves (their existence) do not change. The existence (or non-existence) of the variables are immutable...frozen... static. As Kampis likes to say, it is possible in such non-constructivist dynamics to "shuttle" around in "toy time" (without doing any encoding/decoding) and look at the values on these static variables. The system of differential equations itself (the existence of the variables) is "frozen" at a real world instant. The modeler uses a "toy time" parameter to shuttle through the differential equations frozen at this instance. The term "toy time" is used to express the idea that the system of differential equations may be wholly inadequate at the future time values because the real system may change quite substantially (acquiring new observables that need to be encoded/decoded) during the interval. Because of this, there is no relationship between this "toy time" parameter and real time. All the differential equations represent is a thought experiment of what the real world system would have done if it had behaved exactly like the differential equations in step 2 (not produced or lost anything that needed to be semantically interpreted). What constructivism is: For the constructivists, the important aspect of time is the addition and removal of observables over time. Although it is not necessary to adopt a constructivist stance for something like an electrical circuit (where the electrical components are established {soldered on the board} from the beginning) it is necessary for biological systems which do change their components over time. In constructivism, the modeller stays in the model for all time points in the process. This is necessary so that the encoding/decoding steps can continually be done with the progression of real time. it is only for the special cases of real-world systems that do not experience new (or deleted) observables that the modeler can distance him/herself and let step 2 crank through a certain time interval. In general, for systems that add and delete obsevables, the modeler has to actively participate (continually encoding and decoding the new or deleted observables) for all time points in the time interval of interest. More later. Jeff Prideaux ____________________________________________________________________________________________________________________ From: GEMS::JPRIDEAUX "Jeff Prideaux" 27-FEB-1996 15:59:47.67 Subj: RE: thoughts on constructivism OK, post it!! (My thoughts on constructivism) This too if you like. It actually may address some of the things Seth was asking. The next phase of the story (my thoughts) will be the distinction between deterministic, probabalistic (systems using randomness), and truly creative systems. Component systems, according to Kampis, are creative systems. This makes me feel good because I sometimes think of myself as a creative system...now (at last) there is a body of scientific thought (Rosen and Kampis) that says that creative systems actually exist. Imagine that!! There are many examples...for instance, my dog is also a very creative system (as you can realize by watching her play). My computer, unfortunately, isn't creative. I have to do all the work myself. Ironically, the computer was specifically designed (by very creative and clever people) not to be in the least bit creative. Jeff Prideaux ____________________________________________________________________________________________________________________ Date: Tue, 27 Feb 1996 22:23:34 -0500 (EST) From: Jerry_LR_Chandler Subject: Re: Jeff is cooking. I like when he speaks with his own words!!! Just a few thoughts on Jeff's remarks. A few years ago, I invested a lot of effort trying to understand the nature of a "model" and how it related to "reality." My initial conclusions were similar to what Jeff posted. However, as the real work got started, I found that this distinction between semantics and syntax was insufficient. Eventually, the following schema was devised ( an it has been used with modest success for nearly a decade.) 1. The first step is to establish a classification schema. This is extremely difficult for any system of substantial complexity. 2. Next, do an analysis of the system / data relative to the classification and the existing scientific knowledge base. 3. Next, conduct the synthesis. That is, putting together the knowledge gained from 1 and 2 into a coherent or consistent pattern. If the results of steps 1 and 2 are a set of differential equations, then the problem has blessed you with a degree of cooperation. In biology, the complexity may require that the synthesis be less rigorous - perhaps only a few statistical tests are all that you can get. One may iterate many times between 1,2,and 3. You could view it as cyclic attractor. ;-) 4. Finally, one takes some actions in terms of the goals of the modeling effort. In medicine, this may be prescribing a dosage level, for example. Or, in academic work, publishing the results. The Schema goes under the name "CASA" - classification, analysis, synthesis and action. (So now you know why occasionally my emails come from a machine called "CASA"!) Cheers Jerry ____________________________________________________________________________________________________________________ From: GEMS::WILSONJD "John Wilson - Radiation Biology" 28-FEB-1996 01:03:45.19 Subj: morowitz, hal Don...Hal has favored us with a reply to the following note which I will pass along posthaste. > Dear Professor Morowitz: > > A group of students and faculty members here at the Medical College > of Virginia has been meeting on a regular basis over the last year > to study complex systems. The impetus for establishing the group > has been provided by Don Mikulecky who has browbeaten enough of us > for long enough about the science of complex systems and the > shortcomings of our reductionist ways that it was easier to join > him than beat him. > > We have been spending a great deal of time discussing the work of > Bob Rosen. Recently, in reviewing one of his papers, "Causal > Structures in Brains and Machines" (Int. J. General Systems 12:107, > 1986), a quote attributed to you prompted a spirited discussion > within the group. Your remark addressed the observation that > bacterial cells cooled to absolute zero and subsequently rewarmed, > grow normally, and that the explanation for this in terms of > statistical mechanics is that the velocities are constrained by the > configuration or structure of the system. Some of our discussion > centered on the mechanics of doing this type of experiment and > whether or not your remarks constituted a "thought" experiment or > a real experiment. As it happens, as a posdoctoral student at the > University of Texas with Larry Powers in the late 1960's, I did > exactly these kinds of manipulations with bacterial spores in studies > on radiation sensitivity. I was aware of some of your own earlier > work in this area as well, so when it was suggested that perhaps we > should look at your original paper, I volunteered to track the > reference down. Only then did I discover that the remarks appear > in a 1963 NASA publication (Conference on Theoretical Biology, edited > by G.J. Jacobs) which is not available in our library. In lieu of > trying to locate a copy, I thought I would try to put a couple of > general questions to you via e-mail. First, I presume that your > remarks were not about a theoretical experiment, but that > you were referring to experiments carried out with bacterial spores > cooled to liquid nitrogen or liquid helium temperatures and then > plated out on nutrient medium to assess colony formation. As I > recall, Powers' published some papers with B. megaterium along > these lines in the late 50's. Were you referring to these, or > were you discussing your own experiments? Do you recall what the > theme of the conference was and the general context of the discussion > in which you made your remarks? Your comments would certainly be > appreciated. > > Sincerely, > > John D. Wilson > Associate Professor of Radiology > Division of Physics and Biology ____________________________________________________________________________________________________________________ From: GEMS::WILSONJD "John Wilson - Radiation Biology" 28-FEB-1996 01:09:38.94 Subj: hal II Voila! I have the Artemia ref and will bring it along tomorrow. > In my book "Beginnings of Cellular Life," ( Yale > University Press 1992) the issue you raise is > discussed on pages 52-54. This refers to the > paper Skoultchi and Morowitz Yale J.Biol and Med > 37, 158, 1964 Information Storage and Survival > of Biological Systems Near Absolute Zero. This > work on Artemia eggs at around 2K for 6 days > is convincing evidence. I think that this is im- > portant and I'm delighted that you are discussing > it. Harold Morowitz ____________________________________________________________________________________________________________________ From: GEMS::SBROBERTS 28-FEB-1996 08:12:16.79 Subj: RE: Some thoughts from Jerry Chandler Perhaps Dr. Chandler might be willing to provide us with a simple example (real or imagined) of his CASA scheme? That would be very helpful for thinking about it. ____________________________________________________________________________________________________________________ From: GEMS::JPRIDEAUX "Jeff Prideaux" 28-FEB-1996 08:44:34.68 Subj: C.A.S.A. and constructivism As I understood it, Jerry was saying something like the following: C: classify...or establish a particular context of analysis. Establish a particular way of looking at the system. A: analysis...within this particular context, analyze the system. Identify the functional attributes of the parts that come out of this particular way of interacting with the system. S: synthesis...with the functional attributes identified from the above two steps, form a model of the system. repeat the above three steps as many times as needed (forming larger contexts or ways of looking at the system as necessary) until you have sufficient confidence of the model's adequacy....that the model says something important (you can use) about the real-world system. A: action...make decisions based upon that model. (Or decisions based upon the predictions of that model) -------------------------------------------------------------------------------- I think the only change to the above scenario the constructivists (as described by Kampis) would make is to put the repeat step after the Action. That, in effect, puts the scientist in the system. Kampis would probably say that (for these creative component-systems) we can never form an adequate model (that applies to future time). The best we can do is to continually modify our models, and make decisions along the way. Of course this, in itself, sounds a little pessimistic...although, I'm only half way through Kampis's book. I think one theme he may develop in the later part of the book is that it is possible to make policy decisions based on the knowledge that we cannot have complete knowledge....that knowing that we can't know certain things gives us knowledge. I'll be curious to see what Kampis says in the second part of the book. Jeff Prideaux ____________________________________________________________________________________________________________________ Date: Wed, 28 Feb 1996 13:20:50 -0500 (EST) From: Jerry_LR_Chandler Subject: CASA - and contructivism In-reply-to: DON MIKULECKY > CC: > Subj: C.A.S.A. and constructivism > > You can post this if you like. > > As I understood it, Jerry was saying something like the following: > > C: classify...or establish a particular context of analysis. > Establish a particular way of looking at the system. Three comments: 1. Classification is always for a reason (or purpose, or objective). This reason contributes to the emergence of the context. 2. The term classification was chosen because of it's role in set theory (a set is a class) 3. Philosophically, the term 'Category' was historically used (by Aristotle and Kant, among others) > A: analysis...within this particular context, analyze the system. > Identify the functional attributes of the parts that come > out of this particular way of interacting with the system. Two comments: 1. The root of analysis means to "take apart" Modern usage is extremely diverse, but usually relates to study of the relationships between parts and the whole. 2. Note that theologians and philosophers as well as chemist and mathematicians do "analysis" > S: synthesis...with the functional attributes identified from the above > two steps, form a model of the system. > > repeat the above three steps as many times as needed (forming larger > contexts or ways of looking at the system as necessary) until you have > sufficient confidence of the model's adequacy....that the model says > something important (you can use) about the real-world system. > Two comments: 1. Synthesis literally means "to put together" and is thus the converse (or as "ying is to yang") of analysis. 2. Synthesis in chemistry is a sharply defined and very, very tangible. (The new product is isolated, characterized, structurally defined and shown to be different from its precursors.) As one moves up the scale of complexity, the process of synthesis can become relatively "touchy - feely", that is, the boundaries become blurred between the semantics and the syntax. Big problem. > A: action...make decisions based upon that model. (Or decisions > based upon the predictions of that model) Three comments: 1. Jeff is very perceptive - the goal of this project was to characterize the decision making processes! (In a generic way for scientists and physicians at the FDA working on New Drug applications) 2. When one takes action, one must bring one's human values into play - that is, what Bob Artigiani calls "VEM" - values, ethics and morals. In scientific work, VEM may or may not be important for any particular task. In the FDA, it was a major concern - that is, drug "safety" and the "risk- benefit" decisions developed from the analysis. 3. The individual plays a role in C-A-S as well. Accountability follows more directly from the "action" taken. Issues of durability and stability can strongly influence the nature of the action, almost independent of the synthesis. > -------------------------------------------------------------------------------- > > I think the only change to the above scenario the constructivists (as > described by Kampis) would make is to put the repeat step after the > Action. That, in effect, puts the scientist in the system. Kampis > would probably say that (for these creative component-systems) we > can never form an adequate model (that applies to future time). > The best we can do is to continually modify our models, and make > decisions along the way. I have not read Kampis's book. However, this is where the issues of stability, durability and "lifespan" come in. General speaking, the notion of a "chemical" in endowed with an infinite lifespan unless otherwise noted. (We assume that at absolute zero the half - life for all chemicals would be infinite.) Obviously, this is also where the interface between complexity and evolution enters into the discussion. History continues to emerge. The question becomes, how long the lifespan of a model? In chemistry, our atomic chart has been conceptually complete for roughly 75 years. (Newton's Laws work well for most purposes for over three hundred years.) This should be contrasted with social behavorial patterns for which new "rules" may emerge before I have learned the precursor pattern!! ( I have wandered on a good bit. I hope these contrasts help communicate what the meaning of "CASA". By the way, I eventually learned that the word 'casa' means house in Spanish, so I view the schema as an informal "house of mind") Cheers Jerry ____________________________________________________________________________________________________________________ Date: Wed, 28 Feb 1996 18:26:02 -0400 (EDT) From: DON MIKULECKY Subject: three requests....PLEASE READ I have three requests: 1) Those who were present for Ken Whites super presentation on the application of complexity theory to organizations and professions today, please send in YOUR summarries of what he said! I think that I have been saying all too much and the rest of you to little. Send them to me and I'll post them to the group WITHOUT COMMENT. ____________________________________________________________________________________________________________________ Date: Thu, 29 Feb 1996 09:04 -0500 (EST) From: CARROLLT@PCMAIL.CBIL.VCU.EDU (CARROLL, TIMOTHY J.) Subject: Re: welcome! For me, the most memorable discussion was the elaboration of a distinction between an organization and a profession, especially how the dominant logic (DL) operates differently in each. As I see it, an organization is an integrated administrative structure created to carry out a defined purpose. Selling widgets to make money is a prevalent organizational goal. An organization's DL descibes the process by which it tries to achieve the goal. These DL processes can be changed to better achieve the goal without necessarily altering the identity of the organization itself. A profession, in contrast, is defined by its DL. The shared knowledge, background, ethic, certification, and peer pressure are the glue that binds and provides identity to a profession. These are sociological characteristics, not administrative processes. Consequently, changing the DL of a profession really redefines the profession itself. Everything changes: who belongs, what the members have in common, what society at large expects of the profession. Put another way, an organization's DL determines how it gets things done. A profession's DL determines how it views the world and how the world regards it in return. We discussed the fate of entities that are exposed to a changing environment. They either evolve and adapt, or they die out. If they fail to adapt appropriately, new enities will spontaneously arise to fill the niche left vacant. The specific example given was that in the new health care system RNs may no longer be cost effective. Full time care takers are still needed, however, so someone will come along to fill that role even if RNs refuse to do it on society's terms. I wish we had time to explore the idea that I brought out in the end about the imposition of an outside dominant logic on a profession. As for-profit managed care comes to dominate health care delivery in the U.S., an experiment never before carried out anywhere, how will the new dominant logic of maximizing profits for the shareholders play out? In this scenario, the concept of profession may become irrelevant as the administrative processes goals of the health care organization become paramount. ____________________________________________________________________________________________________________________ Date: Thu, 29 Feb 1996 09:47:19 -0400 (EDT) From: DON MIKULECKY Subject: where are we headed? Professions, Profit and Politics I can't resist the temptation to speculate a bit on the issues Tim brought up about for profit health care and professions. One interesting thing about social structure in the U.S. has been the interplay between the existence of a large middle class and the concept of "professionalism". To some of us, this has always been an interesting myth. The concept of professionalism has many facets. One is the way in which it separates a work force into factions: professional vs non-professional. This division has been viable in part because the professional has often been insulated from the rest of the labor market and thereby has had a priviledged position. In his book "The Iron Heel" Jack London (circa 1920) spelled out a prophecy involving the eventual minimimizing of the middle class by a process in which most of its positions in the work force would be "downgraded" into ordinary labor, and a few would be elevated to a MORE priveledged status. I am wondering if the model we are talking about isn't a sort of manifestation of that prophecy? Why this is especially interesting to me is that Democratic Socialism is based on a "people before profits" value system. London, a socialist, was predicting the onset of a period called "the iron heel" wherein labor was totally at the mercy of the bosses and due to the erosion of the middle class, most of the population was labor. This transition, the movement of large numbers from "professional" status or "executive" or "junior executive" status to an ordinary labor staus had an interesting consequence in London's prophecy. Many of us believe that the "anti-socialist" profit oriented values which have been dominant in this country are a direct result of the willingness of the system to support a large, priveledged middle class. London saw the need to sacrifice this "social buffer" as having a destabilizing result which had to be countered by more control and repression: "the iron heel". In his romantic scenario, after a few hundred years of such repression the people threw it off and a somewhat idealized socialist state replaced it. Yesterday's talk suggests that an alternative model is possible. That by spontaneous adaptive processes, there will be alternative structures which emerge, avoiding the extreme polarization London predicted. I wonder how everyone feels about these ideas? It would seem that some very practical applications of what we are dealing with have been suggested. More of your thoughts will be VERY welcome at this point! ____________________________________________________________________________________________________________________