- NOTES FROM THE EDITOR
- CONFERENCE ANNOUNCEMENT:
- "Artificial Life: A Bridge Towards a New Artificial Intelligence"
- AUTOPOIESIS & ARTIFICIAL LIFE (A-LIFE)
- Part I: An Autopoiesis Simulator
- Part II: Is There Life in C-Life?
- GETTING STARTED WITH AUTOPOIESIS
- AUTOPOIESIS, ONTOLOGY, AND EPISTEMOLOGY: SOME COMMENTS
Now that The Observer is up and running again, I would like to remind you that ANY AND ALL CORRESPONDENCE, COMMENTARY, AND (ESPECIALLY) CONTRIBUTIONS ARE WELCOME. The only thing I ask is that whatever you send be directed toward, closely allied with, or linkable to autopoiesis / enactive cognitive science.
I would also like to remind you that readers are encouraged to send in summary introductions of themselves, their current work, and their interest in autopoiesis / enactive cognitive science.
Finally, any contributions, addenda, or corrections to the ASCII resources (e.g., the autopoiesis bibliography) are welcome. The resources are meant to be just what the label connotes -- shared assets facilitating reading and research in autopoietic theory / enactive cognitive science. -- Randy
THERE HAVE BEEN NO UPDATES TO THE AVAILABLE ASCII RESOURCES DURING THE MONTH OF OCTOBER 1993.
I received the following announcement over internet, and it is of such obvious relevance that I am including it here in The Observer. If you have already seen it -- sorry. -- R.
To: Biology, Engineering and Philosophy Lists
Reposted Message From: Arantza Etxeberria
Subject: Artificial Life Workshop Announcement
Date: Mon., 18 Oct. 93 10:48:42 BST
Palacio de Miramar (San Sebastian, Spain)
December 10th and 11th, 1993
Workshop organised by the Department of Logic and Philosophy of Science, Faculty of Computer Science & Institute of Logic, Cognition, Language and Information (ILCLI) of the University of the Basque Country (UPV/EHU)
Alvaro Moreno (University of the Basque Country)
Francisco Varela (CREA, Paris)
This Workshop will be dedicated to a discussion of the impact of works on Artificial Life in Artificial Intelligence. Artificial Intelligence(AI) has traditionally attempted to study cognition as an abstract phenomenon using formal tools, that is, as a disembodied process that can be grasped through formal operations, independent of the nature of the system that displays it. Cognition appears as an abstract representation of reality. After several decades of research in this direction the field has encountered several problems that have taken it to what many consider a "dead end": difficulties in understanding autonomous and situated agencies, in relating behaviour in a real environment, in studying the nature and evolution of perception, in finding a pragmatic approach to explain the operation of most cognitive capacities such as natural language, context dependent action, etc.
Artificial Life (AL) has recently emerged as a confluence of very different fields trying to study different kinds of phenomena of living systems using computers as a modelling tool, and, at last, trying to artificially (re)produce a living or a population of living systems in real or computational media. Examples of such phenomena are prebiotic systems and their evolution, growth and development, self-reproduction, adaptation to an environment, evolution of ecosystems and natural selection, formation of sensory-motor loops, autonomous robots. Thus, AL is having an impact on classic life sciences but also on the conceptual foundations of AI and new methodological ideas to Cognitive Science.
The aim of this Workshop is to focus on the last two points and to evaluate the influence of the methodology and concepts appearing in AL for the development of a new ideas about cognition that could eventually give birth to a new Artificial Intelligence. Some of the sessions consist on presentations and replies on a specific subject by invited speakers while others will be debates open to all participants in the workshop.
B. Mac Mullin
FOR FURTHER INFORMATION:
E. Mail: firstname.lastname@example.org
Fax: 34 43 311056
Phone: 34 43 310600 (extension 221)
34 43 218000 (extension 209)
Harnad Abstract (others not yet available to poster)
LEVELS OF FUNCTIONAL EQUIVALENCE IN REVERSE BIOENGINEERING: THE DARWINIAN TURING TEST FOR ARTIFICIAL LIFE
Laboratoire Cognition et Mouvement
URA CNRS 1166 I.B.H.O.P.
Universite d'Aix Marseille
II13388 Marseille cedex 13,
Both Artificial Life and Artificial Mind are branches of what Dennett has called "reverse engineering": Ordinary engineering attempts to build systems to meet certain functional specifications; reverse bioengineering attempts to understand how systems that have already been built by the Blind Watchmaker work. Computational modelling(virtual life) can capture the formal principles of life, perhaps predict and explain it completely, but it can no more BE alive than a virtual forest fire can be hot. In itself, a computational model is just an ungrounded symbol system; no matter how closely it matches the properties of what is being modelled, it matches them only formally, with the mediation of an interpretation. Synthetic life is not open to this objection, but it is still an open question how close a functional equivalence is needed in order to capture life. Close enough to fool the Blind Watchmaker is probably close enough, but would that require molecular indistinguishability, and if so, do we really need to go that far?
In The Observer no. 7, Barry McMullin (Dublin City University) posed a number of questions concerning a proposed software simulator for autopoiesis. Even earlier (issue 6) Barry outlined issues of autopoiesis with respect to a TTL logic apparatus titled "Twinkle". Here in issue no. 8, Barry carries on his quest toward analyzing the linkage(s) between autopoiesis and artificial life (a-life).
This section provides some feedback on Barry's questions from issue no. 7.
WHAT LANGUAGES/PLATFORMS SHOULD BE TARGETED? My default inclination is to target the X-windows environment, programming, of course, in ANSI C. A PC-version (under MS-Windows) might be an option also, though further down my list. Comments?
Subscribers to The Observer operate from a variety of computing platforms, if the diversity of their email "bases" is any evidence. Considering that your proposal is for a research tool, a Unix / X-Windows environment would be the odds-on probable favorite. The choice of ANSI C would follow from the Unix preference, but I suppose one must wonder if it would be worth going ahead and doing it in C++. (??) I think Barry's lower / lesser prioritization of MS Windows is a good idea. PC's and other micro's (e.g., the Macs that I and some others use...) can be configured to emulate / tie into an X-Windows environment, thus allowing them "access" to an X-based simulator if need be. The reverse (emulating Windows from a Unix or Mac platform) is not so widely supported by currently available products.
Also, being still an engineer at heart, I wonder if anyone thinks it worthwhile to design a form of hardware accelerator for this class of simulation (conceptually like the CAM cellular automaton machines, but tailored for the needs of autopoietic simulation instead)?
I don't want to sound as if I'm dodging this question, but I don't really see any basis for deciding it until and unless the algorithms and the functionality of the simulator are known. This brings me to what I think is the most critical conundrum:
The Issue of "General Purpose"
The Zeleny APL-autopoiesis software was not really "general purpose", in the sense that (as I understood it) it simulated a restricted set of relations and operations among its units. In other words, a specific set of "game rules" were built into the APL-autopoiesis prototype. To build a "general purpose" simulation platform would require disengaging such specific "rules" or "relations" to the maximum extent possible -- e.g., providing the ability to enter a "rule base" in declarative style. The more succinct way to express this is that a general purpose autopoiesis simulator would be as "domain non-specific" as possible.
This raises the issue of the extent to which this domain-independent rule / relation base facility is to be implemented. At the minimalist end, one could (e.g.) provide little or nothing beyond a set of linkages into the simulator, leaving it for the user to provide a model (domain-specific set of rules / relations) to connect into it. At the maximalist end, one could (e.g.) provide a full-featured graphical user interface, complete with the tools for configuring / specifying the aforementioned model. The minimalist version implies that the simulator might well be deployed as a library of functions to be patched into users' own programs. The maximalist version implies that the simulator might be deployable as a reasonably "stand alone" application. The minimalist version would have limited access (e.g., to C hackers) among the specialized audience of those interested in simulating autopoiesis. The maximalist version would widen the degree of 'access' to those autopoiesis aficionados who are not engineers or computer scientists like Barry and myself, but it would radically increase the level of development work.
A broader issue is that of what functionality would be left once the "domain-specific" aspects were removed. Whatever was being simulated within the "general purpose" simulator would (it seems to me) have to map onto those features of a system which are independent of its specific instantiation. This implies that the distinction between the "general purpose" simulator and the "domain specific" rule / relation base would map onto the distinction between (respectively) 'organization' and 'structure' (within the strict usage of those terms within autopoietic theory).
So what would that leave us with? It would seem that the "general purpose" portion of the simulation package would have to be capable of modeling networks of relations among variable nodes (i.e., 'organization'). This would in turn imply that a reasonable user manipulation / display interface would be graphical, probably arranged in terms of something like directed graphs. "Loading" the simulator with the specification(s) for modeling a particular system (i.e., specifying its 'structure') would then be a matter of programming the reciprocal interplay among relations and states of the structural nodes. This could be very tricky to do (not to mention to provide tools for...), insofar as it would necessarily have to handle recursion.
As some of you may recall, I attempted previously to start a discussion of some foundational ideas in Autopoietic Theory, based on questions framed in terms of Conway's Game-of-Life cellular automaton (C-Life).That discussion got stalled, temporarily at least, as The Observer relocated over the Summer. So now I want to try to re-start the discussion, albeit with one extra twist.
I will not repeat the background here. See, in particular, issue 3 for the original questions, issue 5 for some responses, and issue 6 for my attempted clarification and re-statement, and some further responses.
Note, in particular, that in issue 6 I introduced a definite material realisation of C-life ("Twinkle") to more tightly ground the discussion; as a result I suggest that, in this context at least, a Twinkle-Glider (T-Glider) is certainly a "material system" by any reasonable criterion (I suppose I am still open to argument on this one; but I would prefer to avoid that if at all possible).
The discussion so far has been interesting but, to my mind, has failed to grapple with the really difficult questions. So I now want to try to re-focus the discussion just on what I consider the core issues:
1: Is a T-Glider autopoietic?
2: Is a T-Glider organizationally closed?
I suggest that "Yes" or "No" answers to these can only be defended if you first spell out (as formally as possible) specifically and precisely what you take the "organization" of a T-Glider to be. And any other answers can only be defended by identifying precisely which word(s) in the question(s) are supposed ambiguous or equivocal.
I also add the followed further twist:
3: Is a T-Glider "closed to efficient causation", in the sense introduced by Robert Rosen (Life Itself, Columbia University Press, 1991).
Again, I suggest that a "Yes" or "No" answer can only be defended if you first spell out (yes, formally) specifically and precisely what you take the relational model of a T-Glider to be. And again, other answers would require some precise identification of where the question is ill-defined.
(Of course, one of the directions I am going here is to ask how, if at all, Rosen's "closure to efficient causation" is related to Autopoietic Closure, and/or organizational closure. But for now, I don't want to tackle that issue in its full generality: I'd be very happy to have a secure answer in the relatively narrow context of a T-Glider. So please, try to stick to that narrow context for the time being at least).
'Nuff said. I look forward to any and all responses,
[OK, people --- let's have some comments from the gallery! Barry has put a great deal of effort into presenting the questions he sees as critical to profitable cross-fertilization between autopoietic theory and artificial life. It is our responsibility to match Barry's efforts by joining the discussion. -- Randy]
As you know from issue no. 7, an interactive mailing list on autopoiesis has been started by Kent Palmer (email@example.com). The following is based on something that has come up on the interactive list -- namely, how and where the newcomer should approach autopoiesis / enactive cognitive science. There had been discussion of tutorials and/or introductory materials on autopoiesis. More specifically, there was at least one suggestion that such a tutorial be posted here and/or archived in this vicinity. Although I had addressed this issue in one of the first issues of The Observer (and although it is addressed in the introduction to the bibliography ASCII resource), it just won't go away. Because of the number of new Observer subscribers and the number of subscribers to the interactive list (who can access back issues of The Observer via FTP), I guess it's appropriate to address it once again....
A "standard" tutorial will be difficult to accomplish in any useful fashion. It has been my experience (from having introduced a number of folks to autopoiesis) that:
- IT TAKES A MINIMUM OF TREE OF KNOWLEDGE BEFORE THE NEWCOMER IS CAPABLE OF ASKING RUDIMENTARY QUESTIONS OF THE NATURE: "WHAT THE HECK IS THE CONNECTION TO MY NORMAL WAY OF LOOKING AT THINGS?"
This is the level at which the newcomer can begin to compare and contrast M & V's ideas with respect to his/her own background (e.g., the "representationalism" dominating AI / cognitive psychology). Curiously, this stage of questioning autopoiesis vis a vis the reader's background always seems to occur before that reader starts to question about issues "internal" to the autopoiesis literature. This leads me to the next "stage" I've observed...
- IT TAKES A MINIMUM OF EITHER AUTOPOIESIS & COGNITION OR PRINCIPLES OF BIOLOGICAL AUTONOMY BEFORE THE NEWCOMER IS CAPABLE OF ADDRESSING THE THEORY ON ITS OWN TERMS (MEANING LEXICALLY AND CONCEPTUALLY).
This is the level at which the newcomer is familiar enough with the terminology to ask questions about M & V's ideas on their own (e.g., "organization vs. structure"). I've also noted that it is at this "stage" where newcomers typically begin to see the similarities between autopoietic theory and something they already know.
- THE "RAMP-UP" TIME BEFORE THIS SECOND STAGE IS REACHED IS ON THE ORDER OF 2-3 MONTHS OF REGULAR READING. THE MINIMAL TIME OF 2 MONTHS IS BASED ON MY EXPERIENCE WITH ACADEMICS WITH A STRONG BACKGROUND IN CYBERNETICS AND GENERAL SYSTEMS THEORY.
It takes a while. There's no other way to say it. I've noted that 'hard-core' AI types (cognitivist / objectivists) have a very hard time simply sitting back and absorbing enough of M & V to see what they're talking about. I've also noted that it really helps to have a background in either: (1) philosophy or (2) general systems theory.
I am not claiming that autopoietic theory is incredibly dense or complicated. It does, however, require adopting a new way of looking at things. This shift of perspective is not something that is simply an ordered response to the introductory texts -- it's almost intuitive. The shift comes much easier for those knowledgeable (and comfortable) with general systems theory and/or phenomenology. The shift is much more difficult for those who hold fast to an information-processing / cognitivistic orientation to cognition.
Another complication is that the basic literature is very jumbled. Some of the early stuff was not widely distributed until relatively late; Autopoiesis & Cognition, for example, is a combined re-print of two essays dating from the early '70's. As a book, these didn't appear until 1980 -- a year after Varela's Principles of Biological Autonomy, which was of course based on the material in the 1980 book. (Essential circularity?? -- little autopoietic humor there.. ;-) )
Furthermore, the basic literature is somewhat fragmented or discontinuous. There are some topics or themes which will be found fully treated only in some of the farther-flung papers (e.g., you have to go to a 1978 paper to get Maturana's full analysis of language). There are a number of discontinuities or shifts in terminology, too. The literature trail begins with a 1969 paper by Maturana, but the terminology of "organization", "structure", and "autopoiesis" wasn't in place (print-wise) for another 4 or 5 years.
Finally, the secondary literature is full of people who wish to take the more formalist tenets of this 2nd-order cybernetic theory (one in which the observer is intrinsically embedded) and apply them in a manner more reminiscent of what happened to the "1st-order" cybernetics. By this I mean that the principles originally espoused for living systems get applied to mechanical, software, social, and other complex unities. The literature in which such applications are attempted is often a trap for the newcomer, because M & V's ideas are chopped up, misrepresented, or conveniently ignored. This is not to say that no one but M & V may speak of autopoiesis -- I'm only trying to caution newcomers that some (but not all) of the secondary literature is problematical in a way which is not easily recognized until and unless one has a working knowledge of the basics.
I am therefore very pessimistic about the prospect of setting up a basic tutorial on autopoiesis. Having said all that, let me try to give some pointers.....
Winograd & Flores' Understanding Computers and Cognition (Chapter 4) will give a basic overview. For an even more concise summary, you might want to track down:
Mingers, John, An Introduction to Autopoiesis -- Implications and Applications, Systems Practice, Vol. 2 (1989), no. 2, pp. 159-180.
Mingers, John, The Philosophical Implications of Maturana's Cognitive Theories, Systems Practice, Vol. 3 (1990), no. 6, pp. 569-584.
The first sections of both these papers provide about as succinct and clearly-written summary as I have seen. I would warn, however, that Mingers has a critical agenda of his own, and I neither recommend nor agree with the latter sections of his papers (as regards the needs of a newcomer to the theory).
Moving up the scale, try Tree of Knowledge. This is about as "popular" an account of the theory as will be found.
Finally, when the newcomer feels up to some rigorous training, move on to Autopoiesis & Cognition and/or Principles of Biological Autonomy.
If you must stick with one and only one reference work on a one-shot trial, I would suggest Principles of Biological Autonomy (perhaps excluding the long reprise of George Spencer Brown's Laws of Form), based on its broader topical scope and later date of compilation.
So let me end the tutorial rave here..... My experience is that it takes effort and it takes time. Go to the source, and read it both carefully and with an open mind. -- Randy
Over on the interactive autopoiesis mailing list, a discussion sprang up during October concerning the ontological and epistemological fundamentals for autopoietic theory. In a manner reminiscent of the papers in Zeleny's (1980; 1981) collections, the focus was on where the ontology might be. The following comments are a summary of some of the points I tried to make during the course of that discussion. I would welcome and appreciate any additional comments on this subject. -- Randy
First, autopoietic theory is explicitly epistemological and non-ontological. There has been a persistent push on the part of reviewers and critics to force Maturana and/or Varela to state their ontology, or else to posit one on their behalf. This latter tack -- positing one anyway, when turned around to the point of "positing one that I would like to pursue...", is the basis for a number of positions which apply autopoietic terminology and/or principles to phenomena for which the theory was not targeted ('self-organizing' software phantasms, to name one with popular currency...) or (worse...) phenomena for which Maturana and/or Varela offered caution.
The best example of this concerns the application of autopoiesis to social systems. While a number of writers have attempted to dissect societies with respect to their "organization" and "structure", there is no indication that such an exercise follows from Maturana's and Varela's ideas. The most well-developed example is, of course, Niklas Luhmann, who has built up a vision of societal autopoiesis, but at the expense of adopting "communicative acts" as the fundamental units of analysis. This constitutes a shift of "domains of explanation" away from the biological roots of cognition and toward the network of interactions among observers. Such an analysis affords such "acts" an effective status as ontological primitives. What is problematical is that these new primitives are then treated in isolation from the phenomenology of the living actors who realize them.
Remember -- Maturana started all this as an enquiry into cognition. The theory begins from the viewpoint of how an observer, as a living system, "works". It then proceeds to describe the phenomenology of the living system in action, and (at least for Maturana) how that phenomenology extends to an explanation of interactivity (mutual orientation, structural coupling, etc.).
The entire point is that no explanation for "out there" is valid unless that explanation proceeds from the biological basis of the observer. As such, a strict (and literature-supported) interpretation for Maturana and Varela's work is that there is no ontology exclusive of the observer. The best single reference I can think of for support would be Maturana's 1988 paper in the Irish Journal of Psychology: "Reality: The Search for Objectivity or the Quest for a Compelling Argument". This makes the point pretty clearly. In addition, The Embodied Mind provides a more general overview of the intrinsic links between observer and ontology with respect to philosophy in general and philosophy of (cognitive) science in particular.
For a very good, erudite example of pressure put on M & V to commit to an ontology (in this case, a realist one), see John Mingers' papers in the journal Systems Practice:
Mingers, John, An Introduction to Autopoiesis -- Implications and Applications, Systems Practice, Vol. 2 (1989), no. 2, pp. 159-180.
Mingers, John, The Philosophical Implications of Maturana's Cognitive Theories, Systems Practice, Vol. 3 (1990), no. 6, pp. 569-584.
Mingers, John, The Cognitive Theories of Maturana and Varela, Systems Practice, Vol. 4 (1991), no. 4, pp. 319-338.
Mingers is a very good and cogent writer. He has, however, one major axe to grind --- he can't accept autopoietic theory as a fully-grounded theory until and unless Maturana assents to an ontology, preferably the critical realism of Bhaskar.
To be fair, there is a relative lack of specification on the part of Maturana and Varela as to what they presume the world to be. If one looks closely here and there, the indications are that they proceed from a basis of accepting "realism", but only as a background assumption heavily qualified by the assertion that all theories are but "explanations" (in the sense that term is used in autopoietic theory).
So where does this leave the onto-dependent reader? Going around in circles, it would seem -- specifically the "essential circularity" which Varela et al. start off The Embodied Mind by noting. If the observer cannot disengage herself from explanations derivative from experience / observation, there is no chance of an "objective" ontology coalescing thereby. If the observer were to take those explanations as separate from experience / observation, she would be navigating through territory lying within the range of a number of theoretical and philosophical sources. Autopoietic theory would not be one of them.
Does this make autopoietic theory problematical? You bet it does -- particularly if you have problems dealing with anything other than a linear theoretical progression from a fixed fundament. Such a "linear" progression has been the hallmark of Western philosophy and science -- just plant a (usually 'given') fulcrum somewhere, then obtain leverage by pushing against this fixed point. Whether you plant your fulcrum in the "objective" or the "subjective", it's all the same. One of the promising aspects of cybernetics was its recognition of cyclicity (cf. "feedback"). One of the main reasons cybernetics hasn't (yet) lived up to that promise was the parallel postulation of a new fundament ("information") upon which today's unfortunate objectivist / cognitivist biases were built. For all the holistic, systemic, and processual ideas encompassed by cybernetics, its practice has commonly fallen back into positivistic channels.
Second-order cybernetics was built upon the need to (re-)insert the observer into any system description. A system is what it is, how it is, and where it is (within some hierarchy, perhaps) from some observer's perspective. "Easily circumvented," says the enthusiastic objectivist, "I just blow off the original observer X of system S, and I still see system S!" Refutation of observer-dependency, or a simple transposition of observers? Think about it. As Francisco Varela put it:
"In finding the world as we do, we forget all we did to find it as such, entangled in the strange loop of our actions through our body".
The Creative Circle: Sketches on the Natural History of Circularity, in Watzlawick, P. (ed.) , The Invented Reality, New York: W. W. Norton, 1984, p. 320.
In contrast to linear theoretical progression from the "real", autopoietic theory proceeds from the epistemology as the "primitives", so to speak. This is what seems to turn things topsy-turvy -- the idea that the knowable is dependent on the process of knowing (more specifically, on the knower).
The most problematical aspect of all this is the question (raised by a number of reviewers / critics of Maturana & Varela...) of how one can start with the process of knowing without (at least tacitly) making some commitment to ontological specifics. This apparent lack of ontological foundations is what has proven so irritating to (e.g.) John Mingers. On the other hand, Maturana and Varela, by sticking tightly to a basis in epistemology, have not done anything to assuage the demands of those who can't deal with the issue of knowledge without first circumscribing that which is to be "known".
As I mentioned last week, there are clues here and there throughout M & V's basic writings that they do not dismiss the world or those processes through which living systems are realized (e.g., noting the presumption of physico-chemical laws in Autopoiesis & Cognition).What they do dismiss is any notion of absolute objectivity in the explanations for such phenomena. Mingers (quite reasonably) takes this to be consistent with his preferred realist position, and repeatedly urges that with one small concession (i.e., that there is one unified and objectively extant universe out there, whether or not we can accurately engage and explain it...) he could adopt and adapt autopoietic theory.
To the best of my knowledge, neither Maturana nor Varela have taken Mingers up on his proposal. To my way of thinking, they cannot do so without compromising their theory's postulation of an innately circular and paradoxical nature for cognition. Then again, my "way of thinking", as I am able to reflect upon it intellectively and intuitively, exhibits that very recursive and paradoxical character portrayed in the theory. ;-)
Yes, I would also like to think that there's something fixed about the world. Unfortunately, I can't quite see it for all my flux. ;-) The only way to imbue autopoietic theory with an "ontology" will be to find some way to "fix" the explanation of the observer without respect to her world, or to similarly "fix" the explanation for the world, without respect to any observer.
This way lies madness...... 8-|
End of rave..... End of sermon.....
I apologize for the redundancy, for those of you who have seen more or less this same spiel on the interactive list. I thought it would be useful to send this out to the non-isomorphic population of Observer subscribers in an attempt to open up a philosophical "front" in our discussions. If I have managed to alienate or irritate any of you --- good! Now all you need do is to email your comments for inclusion in the ongoing debate.