Read Darwin's Dangerous Idea Online
Authors: Daniel C. Dennett
20. The evolution of evolvability is a (retrospectively!) obvious recursive move for Darwinians to promulgate—a likely source of cranes, you might say—and it has been One ubiquitous feature in biological evolution that Kauffman concentrates discussed by many thinkers. For an early discussion, see Wimsatt 1981. For a different slant on the issue, see Dawkins 1989b.
on is the principle that "local rules generate global order." This is not 224 BIOLOGY IS ENGINEERING
Stuart Kauffman as MetaEngineer
225
a principle that governs human engineering. Pyramids are always built from signed a special developmental mechanism to protect against such experi-the bottom up, of course, but the organization of the building process, since mentation. (This would be analogous to IBM's forbidding its computer the days of the pharaohs, has been top-down, under the control of a single scientists to investigate alternative architectures for its CPU or central-autocrat who had a clear and literally commanding vision of the whole, but processing-unit chip—
designed
resistance to change.) probably was a bit vague about how the local details would be accomplished.
And what is Kauffman's contrasting explanation? It starts with the same
"Global" direction from on high puts in motion a hierarchical cascade of point and takes it in a rather different direction:
"local" projects. This is such a common feature of large-scale human projects that we have a hard time imagining alternatives (Papert 1993, Dennett
... a locking-in of early development, and hence von Baer's laws, do not 1993a). Since we don't recognize the principle Kauffman discerns as one that represent a special mechanism of developmental canalization, the usual is familiar from human engineering, we are not apt to see it as a principle of sense of which is a buffering of the phenotype against genetic alteration ___
engineering at all, but I suggest that it is. Reformulated slightly, we could put Instead, locking-in of early development is a direct reflection of the fact it as follows. Until you manage to evolve communicating organisms that can that the number of ways to improve organisms by altering early ontogeny form large engineering organizations, you are bound by the following has dwindled faster than the number of ways to improve by altering late Preliminary Design Principle: all global order must be generated by local development. [Kauffman 1993, p. 77. See also Wimsatt 1986.]
rules. So all the early products of design, up to the creation of something with some of the organizational talents of
Homo sapiens,
must obey whatever Think of the issue from the point of view of human engineering for a constraints follow from the "management decision" that all order must be moment. Why is it that the foundations of churches are more alike than their accomplished by local rules. Any "attempts" to create living forms that upper stories? Well, says the traditional Darwinian, they have to be built first, violate this precept will end in immediate failure—or, more accurately, will and any wise contractor will tell you that if you
must
tinker with design not even get started sufficiently to be discernible as attempts.
elements, work on the steeple ornament first, or the windows. You are less If no bell rings, as I have said, to mark the moment when the R-and-D
apt to have a disastrous crash than if you try to come up with a new way of process ends and the life of the "finished product" begins, it should at least preparing the foundation. So it is not so surprising that churches all start out sometimes be hard to tell whether a design principle in question is a principle looking more or less alike, with the big differences emerging in the later of engineering or of meta-engineering. A case in point is Kauffman's (1993, elaborations of the building process. Actually, says Kauffman, there really pp. 75ff.) proposed rederivation of "von Baer's laws" of embryology. One of just aren't as many different
possible
solutions to the foundation problem as there are to later building problems. Even stupid contractors who butted their the most striking patterns in the embryos of animals is the fact that they all heads against this fact for eons would not come up with a wide variety of start out so much the same.
foundation designs. This difference of emphasis may look small, but it has some important implications. Kauffman says we don't need to look for a Thus early fish, frog, chick and human embryos are remarkably similar __
canalization
mechanism
to explain this fact; it will take care of itself. But The familiar explanation for these laws is that mutants [I think he means there is also an underlying agreement between Kauffman and the tradition he
"mutations"] affecting early ontogeny are more disruptive than mutants wants to supplant: there are only so many good ways of building things, affecting late ontogeny. Thus mutants altering early development are less given the starting constraints, and evolution finds them again and again.
likely to accumulate, and early embryos remain more similar from one order of organisms to another than do late embryos. Is this plausible ar-It is the
non-optionality
of these "choices" that Kauffman wants to stress, gument actually so plausible? [Kauffman 1993, p. 75]
and so he and his colleague Brian Goodwin (e.g., 1986) are particularly eager to discredit the powerful image, first made popular by the great French biologists Jacques Monod and Francois Jacob, of Mother Nature as a The traditional Darwinian, on Kauffman's reading, places the responsibility
"tinker," engaging in the sort of tinkering the French call
bricolage.
The term for von Baer's laws in a "special mechanism," built right into organisms.
was first made salient by the anthropologist Claude Levi-Strauss (1966). A Why don't we see many finished products with strikingly different early tinker or
bricoleur
is an opportunistic maker of gadgets, a "sat-isficer"
embryos? Well, since change-orders that affect early parts of the process tend (Simon 1957) who is always ready to settle for mediocrity if it is cheap to be more disastrous in their effect on the finished product than change-enough. A tinker is not a deep thinker. The two elements of classical orders that affect later parts of die process, Mother Nature has de-226 BIOLOGY IS ENGINEERING
Stuart Kauffman as MetaEngineer
227
Darwinism that Monod and Jacob concentrate on are chance on the one hand We have come to think of selection as essentially the only source of order and, on the other, the utter directionlessness and myopia (or blindness ) of the in the biological world. If 'only' is an overstatement, then surely it is watchmaker. But, says Kauffman, "Evolution is not just 'chance caught on the accurate to state that selection is viewed as the overwhelming source of wing.' It is not iust a tinkering of the ad hoc, of bricolage, of contraption. It is order in the biological world. It follows that, in our current view, organ-emergent order honored and honed by selection" (Kauff-man 1993, p. 644).
isms are largely ad hoc solutions to design problems cobbled together by Is he saying the watchmaker
isn't
blind? Of course not. But then what is he selection. It follows that most properties which are widespread in organ-saying? He is saying that there are principles of order that govern the design isms are widespread by virtue of common descent from a tinkered-together process, and that force the tinker's hand. Fine. Even a blind tinker will find ancestor, with selective maintenance of the useful tinkerings. It follows the forced moves; it doesn't take a rocket scientist, as one says. A tinker who that we see organisms as overwhelmingly contingent historical accidents, can't find the forced moves is not worth a tinker's damn, and won't design a abetted by design. [Kauffman 1993, p. 26.]
thing. Kauffman and his colleagues have made an interesting set of discoveries, but the attack on the image of the tinker is to a large extent, I Kauffman wants to stress that the biological world is much more a world think, misplaced. The tinker, says Levi-Strauss, is willing to be guided by the of Newtonian discoveries (such as Turing's) than Shakespearean creations, nature of the material, whereas the engineer wants the material to be perfectly and he has certainly found some excellent demonstrations to back up his malleable—like the concrete so beloved by the Bau-haus architects. So the claim. But I fear that his attack on the metaphor of the tinker feeds the tinker is a deep thinker after all, complying with constraints, not fighting yearning of those who don't appreciate Darwin's dangerous idea; it gives them. The truly wise engineer works not
contra naturam
but
secundum
them a false hope that they are seeing not the forced hand of the tinker but
naturam.
the divine hand of God in the workings of nature.
One of the virtues in Kauffman's attack is that it draws attention to an Kauffman himself has called what he is doing the quest for "the physics of underappreciated possibility, one that we can make vivid with the help of an biology" (Lewin 1992, p. 43), and that is not really in conflict with what I am imaginary example from human engineering. Suppose that the Acme Ham-calling it: meta-engineering. It is the investigation of the most general mer Company discovers that the new hammers made by its rival, Bulldog constraints on the processes that can lead to the creation and reproduction of Hammer, Inc., have plastic handles with exactly the same intricate pattern of designed things. But when he declares this a quest for "laws," he feeds the colored whorls on them as is sported by the new Acme Model Zeta. "Theft!"
antiengineering prejudice (or you might call it "physics envy") that distorts so scream their legal representatives. "You copied our design!" Maybe, but then much philosophical thinking about biology.
again, maybe not. It just might be that there is only one way of making plastic Does anyone suppose that there are
laws
of nutrition? Laws of locomotion?
handles with any strength, and that is to stir up the plastic somehow as it sets.
There are all sorts of highly imperturbable boundary conditions on nutrition The result is inevitably a distinctive pattern of whorls. It would be almost and locomotion, owing to fundamental laws of physics, and there are plenty impossible to make a serviceable plastic hammer handle that
didn't
have of regularities, rules of thumb, trade-offs, and the like that are encountered by those whorls in it, and the discovery of this fact might be one that would be any nutritional or locomotive mechanisms. But these are not laws. They are eventually imposed on just about anybody who tried to make a plastic like the highly robust regularities of automotive engineering. Consider the hammer-handle. This could explain the otherwise suspicious similarity regularity that
{ceteris paribus)
ignition is accomplished only by or after the without any hypothesis of "descent" or copying. Now, maybe the Bulldog use of a key. There is a reason for this, of course, and it has to do with the people did copy Acme's design,
but they would have found it in any case,
perceived value of automobiles, their susceptibility to theft, the cost-effective sooner or later. Kauffman points out that biologists tend to overlook this sort (but not foolproof) options provided by preexisting locksmith technology, and of possibility when they draw their inferences about descent, and he draws so forth. When one understands the myriad cost-benefit trade-offs of the attention to many compelling cases in the biological world in which similarity design decisions that go into creating automobiles, one appreciates this of pattern has nothing to do with descent. (The most striking cases he regularity. It is not any kind of law; it is a regularity that tends to settle out of discusses are illuminated by Turing's 1952 work on the mathematical analysis a complex set of competing
desiderata
(otherwise known as norms). These of the creation of spatial patterning in morphogenesis. ) highly reliable, norm-tracking generalizations are not laws of automotive In a world with no discoverable principles of design, all similarities are engineering, nor are their biological counterparts laws of locomotion or suspicious—likely to be due to copying (plagiarism or descent).
nutrition. The location of the mouth at the bow rather than the stern end of the locomoting organism (
ceteris paribus
— there are exceptions!) is a deep regularity, but why call it a law? We under-
228 BIOLOGY IS ENGINEERING
stand
why
it should be so, because we see what mouths—or locks and keys—
are
for,
and why certain ways are the best ways of accomplishing those ends.
CHAPTER NINE
CHAPTER 8:
Biology is not fust like engineering; it
is
engineering. It is the
study of functional mechanisms, their design, construction, and operation.
Searching for Quality
From this vantage point, we can explain the gradual birth of function, and the
concomitant birth of meaning or intentionality. Achievements that at first
seem either literally miraculous (e.g., the creation of recipe-readers where
none were before) or at least intrinsically Mind-dependent (learning to play
winning checkers) can be broken down into die ever smaller achievements of
ever smaller and stupider mechanisms. We have now begun to pay close
attention to the design process itself, not just its products, and this new
research direction is deepening Darwin's dangerous idea, not overthrowing
it.