Review of: "Essays on Life Itself" by Robert Rosen, Columbia University Press, (estimated date of publication, November 1999).
Donald C. Mikulecky
Professor of Physiology
Medical College of Virginia Commonwealth University
Box 980551 MCV Station
Richmond, VA 23298-0551
Robert Rosen died in December of 1998 after a long bout with diabetes and its complications. He left a significant quantity of unpublished notes and had this book in the publication process. His last "writings" were hand done on paper with great effort due to extensive peripheral neuropathy. It was a mixed blessing to be among the first to read his last works both this manuscript and the next, unfinished one. I am saddened by our loss even as I feel his presence through his writings.
Bob was an eloquent speaker and reading this set of essays is almost as good as hearing him in person. The essays were written to be published in a number of places, usually as invited talks, yet they may as well have been set down to be a book from the start. There is a thread of continuity that makes this the case. In addition, even though I had read many of the essays as they appeared earlier, their juxtaposition in this volume proves that "the whole is more than the sum of its parts"!
The book is 344 pages. The parts come as five, each containing chapters within (the individual essays). They are:
PART I: On Biology and physics
PART II: On Biology and the mind
PART III: On genericity
PART IV: Similarity and dissimilarity in biology.
PART V: On biology and technology.
His stated purpose of this collection is to, in a sense, "flesh out" arguments in Life Itself (LI) that had to be short or even omitted for what might be called "logistic" reasons. In my opinion they essays do that at least. In LI he began with a caveat with which I am totally sympathetic. He warned the reader that he was weaving a very intricate cloth with a single linear thread and therefore much was being laid upon the reader's shoulders. My own experience is that it took numerous readings to begin to see how the weave was manifest. Once there, things fell into place more and more quickly, yet still a lot more was required because the design is so highly interconnected and rich in levels of meaning. I hope this book of essays will spare others that struggle. It will never be my place to evaluate that possibility since I can never go back.
For some years I have maintained a pair of web sites mainly devoted to these ideas. They are at http://views.vcu.edu/complex/ and the site listed in the heading. In addition, I ran a discussion group from the complexity site and participated in two others, the Principia Cybernetica Project (PCP) and the newer New England Complex Systems Institute (NECSI) both of which are linked from the complexity site. There is no way I can relate how much my present state of mind and my understanding of Rosen's work has been influenced by exchanges on these boards. We are truly in a new age when such things are possible.
In addition, since Bob Rosen was to have participated in the 43rd annual International Society for the Study of Systems meeting in 1999, I was asked to come to summarize his work in his absence. That review appears on my web site as well as on theirs and in the proceedings. So far those giving feedback have found it helpful. For that reason I will refer to it rather than rewrite a lot of background material here.
The first part deals with the relationship of biology and physics within science, which can sound like an innocent enough topic until one understands that it is a revolutionary view.
It is revolutionary because of the depth to which Rosen delves to get at the root of a number of matters (and already we have to cope with the intertwined threads of thought).
Underlying it all is the common notion that physics is the source of all scientific laws and that chemistry and biology somehow must utilize physics to be scientific. Rosen rejects this notion and thereby opens a Pandora's Box. He uses the now fifty-five-year-old essay by Schrödinger, What is Life? as a springboard to the revealing argument about biology's more generic character in comparison to physics. As he does this he develops his notion of complexity as a description of this more generic view promoted by biology in contrast to the kind of "simple systems" which are the subject matter of physics. None of this should sound new to anyone who has read his earlier work, especially Life itself, except for the new connections and new depths to which the arguments are taken. The result is a more solid whole than ever before.
I will mention a few lines of argument that particularly impressed me to try to entice everyone to read the whole thing. The first essay is entitled "The Schrödinger Question, What is Life?" and deals with what Rosen considers the basic question of biology. His introduction to this part of the book is worth having here to get a flavor for where he is going: "I claim that Gödelian noncomputability results are a symptom, arising within mathematics itself, indicating that we are trying to solve problems in too limited a universe of discourse." This is a nice capsule version of Rosen's message. If nothing else comes from his writings, this alone should change everyone who understands the message.
Much of the book is devoted to explaining why this is so. This chapter looks into Schrödinger's essay to find clues to the relation of the answer to the question to the demonstration that the methods offered by physics are too non-generic to answer the question.
A line of reasoning running through all his writings is the ease with which things that seem to be the domain of physics and physics alone can be recast into biological terms. One set of concepts that gets used this way is the relation of genotype to phenotype. The careful examination of this relationship weaves into the basic failure of reductionism to supply the whole from its parts. This, of course, is just another way of saying that the simple systems of physics fail to describe the complexity of the real world.
Another approach to this distinction between simple and complex comes from a comparison of inertial and gravitational forces and the resulting issue of the difference between response to a force and the generation of a force. He then looks at the gene as a molecule argument from this perspective and reminds us that when such an identification is made it shifts the object from a "gravitational" role to an "inertial" role! This is but one example of his ability to use analogy and generalization to weave the picture. As it emerges, we see the connection between a sparseness of meaning in physics, the need to go outside physics (in this case to biology) to augment the formalism, and the epistemological ramifications of looking beyond the usual limited usage of these words. He then turns around the popular and familiar idea that the discovery that the gene can be a molecule was what matters most. He asserts that this obscures the more important question: "When can a molecule be a gene?" Finally, he weaves in another line of reasoning which further illuminates the point being made. He discusses certain structures with a grammatical analog. He compares "life" as a noun with "living" as an adjective and generalizes that comparison to graphical patterns of causality, which are central parts of his treatment. This essay is well placed as it serves as a good preview and enticement for what is coming.
The second essay in this part, "Biological Challenges to Contemporary Paradigms of Physics and Mimetics" are aimed at two problems, first the role of physics in dealing with the properties of living systems, and second, the role of mimetic approaches (prefixed by "artificial"). In this essay he tackles reduction of biology to physics head on. Here he weaves in another thread, the ontology of the "something" that makes the whole more than the mere sum of its parts (and, generally, different from its parts). Here is n essence that has to be appreciated if one is to appreciate Rosen at all. There is actually something here which gets lost when the system is fragmented. It is non-fragmentable. How do we "see" such things? We observe what the system does. We see that disappear when the fragmentation is performed. These things are defined by their context. They are context dependent! This idea now takes on a life of its own feeding on the other threads in the weave much as what it is telling us about. He then contrasts this with the strong desire to obtain objectivity through context independence in physics. He weaves back in the non-generic character of such things. He then shows the inability of mimesis to cope with these issues and weaves back in the thought that complex systems are not simulable.
In the third chapter, "What is biology?" he goes further in his comparison of Mendelian genes, phenotypes and DNA. He considers three possibilities, based on the fact that simple systems posses a "largest model" while complex system do not.
The first recognizes that DNA and Mendelian factors are not readily matched but asks that somehow there is a way of getting from one to the other. The second says that the first is not possible, but that a largest model can be used to obtain them independent of that. Finally, we must consider that there is no largest model and that the system is complex.
The next part deals with the biology of the mind. In the introduction he states that the chapters in this part deal with "objectivity". He asserts "As a whole, they argue that attempts to identify objectivity with algorithms and their execution has been a profound mistake.
The first chapter of this section, chapter 4, "The Church-Pythagorous Thesis" reviews some 2000 years of mathematical history to trace Church's Thesis to Pythagorous. It culminates in the recent failed attempts to reduce mathematics to algorithms. In so doing he makes his refutation to the claim that all effective processes are computable. The section on Pythagorous is delightful. It weaves together the issue of genericity with the issues surrounding measurement and the limited use of algorithms to do even basic mathematics. It also raises the issue of commensurability and its devastating effect on the formalist's efforts. The role of mimesis can be summed up by the conclusion that mimesis is not science, it merely mimics science.
Chapter 5, "Drawing the Boundary between Subject and Object: Comments on the Mind-Brain problem", is a paper published in the Journal of Theoretical Medicine. This paper is an example of its own subject and therefore exemplary of the notion of self-referential loops and impredicativities. This of course is a further example of the context dependence of complex systems in contrast to simple systems. In this chapter he reviews the use of mathematics in the study of mind and sets the record straight about the history of neural network theory that actually started with the early work of Rashevsky. This is his first explicit attempt to talk about the mind-brain problem and it immediately is related to the "measurement problem" in physics and the whole basis for the concept of "objectivity". He says in the introduction to this part of the book: "Hence, by extension, a reductionist excursion into the mind-brain problem in the name of objectivity merely exposes the limitations of the reductionist approach itself."
Chapter 6, "Mind as Phenotype" weaves the mind-brain problem into the existing cloth by using a generalization of the arguments already presented in earlier chapters. It is a sequel to the previous chapter, developing a new angle. Here, another important thread is woven in, the Aristotelian causalities. The genome is identified as formal cause and related to phenotype in this manner. The question of the legitimacy of this approach is examined. The inadequate roles of reductionist fractionation and mimesis are dealt with as the examination proceeds. The following admonition is worth noting: "An essential feature of this discussion is the necessity of learning about a material system, not only by taking it apart into subsystems, but also by putting it into larger systems with which it can interact. This is still another set of threads based on Gödel and leading to the notion that complexity demands going outside the system for answers.
This may be a good place to interject a few thoughts from outside the system that Rosen has created. One is the overlap between Rosen's concept of organism and Maturana and Varela's autopoiesis. Another is the use of the modeling relation and semiotics. These are two areas where what Rosen and others are saying may interact in a very constructive way.
Rosen's answer to this is the relational approach that is quite compatible with impredicativites and reaching outside the system for additional system definition due to its context dependence.
Chapter 7, " On Psychomimesis" is a further discussion on mimesis and, in particular, the Turing test. The contrast between mimesis and science is a contrast between synthetic and analytical approaches to the same end. If analysis and synthesis, as used in this context, were simple inverses they would be reconcilable. In complex systems they are not. In short, the mimicry uses a different set of causal relation than exist in the system being imitated in order to obtain a certain form of behavior. The syntactic methods used are unable to deal with the causal relations in the system being examined and thereby circumvent the entire issue. Imitation substitutes for understanding.
Chapter 8, "The Mind-Brain Problem and the Physics of Reductionism" is a direct assault on reductionism and its failure to address the issues it has been used to address. This is because of the absence of algorithms and/or lists of properties which somehow go together to make the whole system.
In this chapter, the essence of what is lacking in the traditional approach to science is summed up very nicely in the critique of the study of the mind: "On the other hand, experience has shown that the resultant lists or algorithms either turn out to be infinitely long, which is unacceptable, or else must turn back on themselves, which is also unacceptable. The latter essentially constitutes impredicativity - Bertrand Russell's viscous circle. Such impredicativities cause semantic referents within them, in this case self-referents that depend entirely on the context created by the circle itself. Attempts to eliminate such circles by purely syntactic(context-independent) means, to express them as a finite list, even in a bigger syntactic system, simply destroy these properties."
Part III of the book is about Genericity. It expands on chapter 2 of Life Itself which dealt with what is general and what is special. He uses Rene Thom's work on Structural Stability and morphogenesis as an example. He then generalizes the concepts by talking about comparing things with their nearby neighbors with respect to some property. If the nearby neighbors also posses the property it is called generic.
Chapter 9, "Genericity as Information", was an essay he contributed to a Festschrift for Rene Thom and it will appear there as a French translation. It harks back to the "What is Life" question and goes further into what can be learned from the interchange of noun and adjective. This thread has much to offer. For example, the words turbulent and turbulence, illustrate the issue very well. As an adjective describing water, air, or oil we find the usual usage that science is comfortable with. It describes a property of real substances. Turn it around and talk about water turbulence, oil turbulence, and air turbulence and the quality has become the object of study rather than the material. The same is true with life and living. Once life becomes a thing, an object of study, science looses its ability to prescribe an approach. We have no process that can separate the noun and adjective aspects of such qualities, they are outside the paradigm. This raises the issue of surrogacy, a thread that intertwines with the modeling relation and other ideas. Catabolic and anabolic modes are clearly aspects of the organism. They are an essential part of the M-R systems and their role in defining life. Reductionism reduces life to a "pile of rubble" and asks that this be a surrogate for the whole, losing any trace of these and other essential properties. Had the issue of genericity been understood, such a poor surrogate would have never been considered.
Chapter 10, "Syntactics and Semantics in Languages, deals with a central thread of the complexity theme. The duality between syntax and semantics is the key to the impossibility of ever bringing reality into a simple system mode of description without the loss of most, if not all, of the meaning. Rosen uses mathematics, a particular language, as a surrogate for languages in general to make his point. He reviews the failure of formalization and its roots in Gödel's work. Church's Thesis is again rejected and the role of inferential structure through entailments is studied. Many of the threads introduced earlier are thereby given further credibility. The modeling relation is reviewed as a means of illustrating how these ideas come together. The recent work of Dress strengthens this tie very well . It puts semiotics into the form of the modeling relation and weaves that together. The "well posed problem" is discussed. The idea that this means stripping off semantics and reducing the problem to syntax suggests that the well posed problem discussion itself is not well posed. Church's Thesis is given more attention using the newly developed threads along with those already woven. It is held up as an antithesis to the well-posed problem. The reliance of complex systems on external referents and closed loops internally makes a strong bond between language and the rest of the world, the external referents of language.
Chapter 11, "How Universal is a Universal Unfolding?" is a more technical, mathematical chapter. In it, Rosen weaves in the argument that involves the exactness of differential forms and the absence of state descriptions of complex systems. Having included an appendix about Carotheodry's proof of the Second Law of Thermodynamics using the integrability of differential forms in my own book, I find this argument a very stimulating one. It has a distinct non-mechanistic, thermodynamic flavor. Here Rosen also weaves the thread further with the distinction between anabolic and catabolic processes and the fact that they are not inverses. This thread delves deeply into the nature of the organism vs. machine distinction along the lines of what is called "path dependence" in the categorization of work and heat as something other than state functions in thermodynamics. What Rosen shows us is that the work and heat type variable is more generic than the state function! The thread about catabolism and anabolism is developed further with the recognition that failure modes, such as the breaking of the beam, are not the reverse of manufacture or synthesis.
Chapter 12, "System Closure and Dynamical Degeneracy" follows up along the lines of the previous chapter. It begins with the dissociation of degeneracy from genericity. Degeneracies are generally not generic. This might seem trite until we recognize how much of what we pay attention to in science is a degeneracy. Even the numbers we use most often are degenaracies among numbers. The discussion turns to the openness of living systems and the degeneracy of closed systems. Alternative views of senescence are couched in this language giving the phenomenon a very fresh meaning.
Chapter 13, "Some Random Thoughts about Chaos and Some Chaotic Thoughts About "Randomness" is a chapter which helped inaugurate the Journal of Biological Systems. I was so impressed by this essay that I used it as a basis of my own essay, " A Close Look at a New Science: Chaos as Science or Science in Chaos" In this essay, the genericity of strange attractors and notions like "life at the edge of chaos" are put into a perspective commensurate with the rest of Rosen's thinking. He is careful to avoid getting into specific arguments about the application of chaotics to science saying it is much too early to tell how much it might eventually contribute. It is clear that he sees a potential for hype and fad to cloud the real nature of what chaos is. In its most clear form it is merely a class of solutions to equations of motion within non-linear dynamics. He also reminds us that as a model for real systems, it has been kept in a "hot house" without environmental or other interactions. There are also discrete forms and continuous forms of equations exhibiting chaotic solutions. Meanwhile, the role of randomness as a model for those parts of systems that we do not intend to specify in detail seems to still have a lot of merit.
In Part IV that deals with similarity and dissimilarity in biology, Rosen covers a number of topics including biological form and morphogenesis. Chapter 14, "Optimality in Biology and Medicine" was originally a tribute to Richard Bellman who was the creator of dynamic programming, a powerful optimization technique.. Rosen has written his own book on optimality in biology. One of the most striking threads is woven in here. He shows that by the use of Hamilton's principle of least action and the analogy from its application in optics, it should have been possible to develop the Schrödinger equation for wave mechanics much earlier. This is a kind of object lesson that the strict adherence to mechanistic thinking in the Newtonian Paradigm can actually keep us from insights gained by other, parallel thought patterns, such as the one being developed in this volume.
Chapter 15, "Morphogenesis in Networks", is Rosen's own version of what is now generally called "Artificial Life". My phrasing it this way is deliberately meant to be provocative because he never thought any of the large body of work called that should have ever been so named. In spite of that, the chapter itself views these objects in a fresh way and makes use of the other ideas in this volume to give the work a different perspective from those who claim too much for a mimetic approach. This chapter would be a good one to look at if one were looking for a thesis topic. It would be even more suggestive if the notion of causal entailment networks in his unpublished work were more readily available.
Chapter 16, "Order and Disorder in Biological Control Systems", is motivated by the problem of senescence. This also is an area where Rosen saw new horizons, yet never was able to complete the journey. He began the ideas in a few papers and in the book Anticipatory systems. His unpublished work carries it a bit further, especially along the lines begun in the earlier works that deal with the whole notion of time. He speaks in terms of the links to the future in the way causal relations are set up in networks. He then introduces the concept of a general system failure as a holistic view of what mechanistic reasoning has failed to explain in spite of much effort and expense. Finally, another important thread is woven in. The existence of side effects in a system captures in still another way the essence of what complexity is all about and why the machine metaphor is inadequate.
Chapter 17, "What Does it Take to Make an Organism?", is an extremely provocative chapter. In it Rosen reviews the development of his concept of organism as distinct from machine using the causal entailment network. He then goes further than ever before toward revealing some things he discussed in a taped interview in July of 1977. In that interview he discussed the moral implications of revealing what he knew about the possibility of fabricating organisms. He was reluctant to do so, yet also felt a pressure to do so. I am not aware of whether or not more information exists on this topic, but I will say that this chapter has to be read by anyone even remotely concerned with the issue.
The final part of the book, Part V, on biology and technology is a treat in its own right. Once again if one wants ideas for future work, this is a rich source. The focus is on what biology has to teach us about other disciplines. Rosen believed that biology was a means for learning to approach complex problems, especially those in human society. The quickness with which complexity science has been applied to human organizations and institutions says that he was on the mark this time too. These matters were reluctantly omitted from life itself for logistic reasons. He says: "The common relational models that bridge biology and the technologies allow us, in principle, to separate the fruits of selection without needing to emulate its methods. They provide a Rosetta stone that allows us to utilize the billions of years of biological experience contained in Nature's encyclopedia, and to realize them in our own ways, applied to our own problems." What a challenge to those who are looking to the future and have realized the latent power of biotechnology!
In this section the metaphor of chimera is introduced as a new thread and woven in with the rest. It connotes a single organism with more than the usual number of parents. Its cells arise from genetically diverse sources. He then makes the analogy with social structures and institutions. He takes the analogy in the other direction to speak about activated complexes in biology. He says " One of the deepest lessons of biology is that such cooperation [between function in chimeras] is selected for; indeed, that life would be impossible with it; and hence that complex organizational problems can be solved via cooperation and not by power and competition."
Chapter 18, "Some lessons of Biology", begins with the biggest lesson to be learned from biology, namely that there are lessons to be learned from biology! The discussion revolves around the hermit crab that becomes an immediate symbol for the cloth woven so far in many ways. The special chimerical nature of this crab in a shell with an anemone attached is full of meaning that he exploits in a delightful way. He concludes that life is too rich a process to rely on programs for its source. Therein is one big clue to the fabrication question.
Chapter 19, "Bionics Revisited" might somehow use the word "cybernetics" to advantage, but doesnt. beginning with some historical remarks, Rosen develops a strategy for biotechnology worth considering seriously. Using the bird wing as a talking point he once more shows us the futility of holding too tightly to the machine metaphor. He then weaves to a conclusion his ideas about complexity, pointing the way to the future and what might be done.
Chapter 20, "On the Philosophy of Craft", is a delightful excursion into the world of reality where "magic bullets (treatments without side effects) do not exist. He then proceeds to tell us what can be expected in such a world. He says:"We are now free to envision strategies involving controllers that are themselves complex .in fact for complex systems in general, complex controllers are the only reasonable hope."
Chapter 21, "Cooperation and Chimera" and Chapter 22, "Are Our modeling Paradigms Nongeneric?" deal with an approach to complex systems based on dynamics. These are in some way the main thrust of the volume. Let me end this summary with a quote from the introduction of this section: " the fabrication of something(e.g., an organism is a vastly different thing than the simulation of its behaviors. In conclusion, any material realization of the (M<R) system [model for an organism] must have noncomputable models ..Thus we have part of the answer to the question with which we started---What does it take to build an organism? It takes a lot more than we presently have. That is why the problem is so hard, but also why it is so instructive."
In conclusion, this book of essays is unique in its scope and content. The scientific world is beginning to discover Robert Rosen, not so much because it wants to or because it likes what he teaches, but because there may be no other way to proceed.
Back to Mikulecky's home page
Back to Complexity Home Page