Read Darwin's Dangerous Idea Online
Authors: Daniel C. Dennett
Moreover, much that appears unremarkable when you look at a cathedral as simply a finished product seems deeply puzzling when you start asking 18. The preceding three paragraphs are drawn, with revisions, from Dennett 1994a.
how it could have been built. Chicken-egg problems abound. If you build the flying buttresses before you build the central vault, they will push the 218 BIOLOGY IS ENGINEERING
walls in; if you build the vault first, it will spread the walls before the buttresses can be installed; if you try to build them both at once, it seems likely that the staging for one would get in the way of the staging for the other. It is surely a problem that has a solution—probably many different ones—but thinking them up and then looking for the evidence to confirm or disconfirm them is a challenging exercise. One strategy that recurs is one we have already seen in action in Cairns-Smith's clay-crystal hypothesis: there must have been scaffolding members that have disappeared, that functioned only during the building process. Such structures often leave clues of their former presence. Plugged "putlog holes" are the most obvious. Heavy timbers called "putlogs" were temporarily fixed in the walls to bear the scaffolding above them.
Many of the decorative elements of Gothic architecture, such as the elaborate patterns of ribs in the vaults, are really structurally functional members—but only during the construction phase. They had to be erected before the "web courses" of the vaults could be filled in between them. They stiffened the relatively delicate wooden "centering" scaffolding, which would otherwise have tended to buckle and deform under the temporarily uneven weight of partially built vaults. There were severe limits on the strength of scaffolding that could be constructed and held securely in place at great heights using medieval materials and methods. These limits dictated many of the "ornamental" details of the finished church. Another way of making the same point: many readily
conceivable
finished products were simply impossible to erect, given the constraints on the building process, and many of the apparently non-functional features of existing buildings are in fact enabling design features without which the finished product could not exist.
The invention of cranes (real cranes) and their kin opened up regions of the space of architectural possibility that were previously inaccessible.19
The point is simple, but casts a long shadow: When you ask functional questions about
anything
—organism or artifact—you must remember that it has to come into its current or final form by a process that has its own requirements, and these are exactly as amenable to functional analysis as any features of the end state. No bell rings to mark the end of building and the beginning of functioning (cf. Fodor 1987, p. 103). The requirement that an organism be a going concern at every stage of its life places iron constraints on its later features.
19. Foui classic explorations of these issues are John Fitchen's
The Construction of
Gothic Cathedrals,
which reads like a detective story, Fitchen's
Building Construction
Before Mechanization
( 1986), William Barclay Parsons'
Engineers and Engineering in
the Renaissance
( 1939, republished by MIT Press, 1967) and Bertrand Gille's
Engineers
FIGURE 8.1. Early rotating cranes and other
of the Renaissance
( 1966 ).
devices for raising or moving loads. (From Diderot and d'Alembert,
Encyclopedic
[1751-1772], reproduced in Fitchen 1986.)
220 BIOLOGY IS ENGINEERING
Stuart Kauffman as Meta-Engineer
221
D'Arcy Thompson (1917) famously said that everything is what it is brand-new science: chaos theory and complexity theory, strange attractors because it got that way, and his own reflections on the historical processes of and fractals. He himself has been tempted by that view in the past (Lewin development led to his promulgation of "laws of form" that are often cited as 1992, pp. 40-43), but his book bristles with warnings, fending off the examples of biological laws that are irreducible to physical laws. The embrace of the anti-Darwinians. He begins the preface of his book (p. vii) by importance of such reconstructions of developmental processes and the describing it as "an attempt to include Darwinism in a broader context": investigation of their implications is undeniable, but this issue is sometimes misplaced in discussions that attempt to
contrast
such developmental Yet our task is not only to explore the sources of order which may lie constraints with functional analyses. No sound functional analysis is com-available to evolution. We must also integrate such knowledge with the plete until it has confirmed ( as much as these points ever can be confirmed ) basic insight offered by Darwin. Natural selection, whatever our doubt in that a building path has been specified. If some biologists have habitually detailed cases, is surely a preeminent force in evolution. Therefore, to overlooked this requirement, they are making the same mistake as the art combine the themes of self-organization
and
selection, we must expand historians who ignore the building process of their monuments. Far from evolutionary theory so that it stands on a broader foundation and then being too taken with an engineering mentality, they have not taken engi-raise a new edifice. [Kauffman 1993, p. xiv.]
neering questions seriously enough.
I go to such lengths to quote Kauffman himself on this point since I have also felt the strong wind of anti-Darwinian sentiment among my own readers 7. STUART KAUFFMAN AS META-ENGINEER
and critics, and know that they will be strongly motivated to suspect that I am merely reworking Kauffman's ideas to fit my own biased view! No, he
Since Darwin, we have come to think of organisms as tinkered-together
himself—for whatever that is worth—now sees his work as a deepening of
contraptions and selection as the sole source of order. Yet Darwin could
Darwinism, not an overthrow. But, then, what can be his point about "spon-not have begun to suspect the power of self-organization. We must seek taneous self-organization" as a source of "order" if not a flat denial that
our principles of adaptation in complex systems anew.
selection is the ultimate source of order?
Now that it is possible to build truly complex evolutionary scenarios on
—STUART KAUFFMAN, quoted in Ruthen 1993, p.
138
computers, rewinding the tape over and over, we can see patterns that eluded earlier Darwinian theorists. What we see, Kauffman claims, is that order History tends to repeat itself. Today we all recognize that the rediscovery
"shines through"
in spite of
selection, not because of it. Instead of witnessing of Mendel's laws, and with them the concept of the gene as a unit of heredity, the gradual accrual of organization under the steady pressure of cumulative was the salvation of Darwinian thinking, but that was not how it appeared at selection, we witness the
inability
of selective pressure (which can be the time. As Maynard Smith notes (1982, p. 3), "The first impact of carefully manipulated and monitored in the simulations) to overcome an Mendelism on evolutionary biology was distinctly odd. The early Men-inherent tendency of the populations in question to resolve themselves into delians saw themselves as anti-Darwinians." This was just one of many self-ordered patterns. So this seems at first to be a striking demonstration that styled anti-Darwinian revolutions that have turned out to be pro-Darwinian natural selection cannot be the source of organization and order after all—
reformations, dragging Darwin's dangerous idea from one sickbed or another which would indeed be the downfall of the Darwinian idea.
and putting it back to work. Another that is unfolding before our eyes today is But there is another way of looking at it, as we have seen. What conditions the new direction in evolutionary thinking spearheaded by Stuart Kauffman have to be in effect for evolution by natural selection to occur? The words 1
and his colleagues at the Santa Fe Institute. Like every good bandwagon, it has put into Darwin's mouth were simple: Give me Order, and time, and I will a slogan: "Evolution on the Edge of Chaos." Kauffman's new book,
The
give you Design. But what we have subsequently learned is that not every
Origins of Order: Self-Organization and Selection in Evolution
(1993), variety of Order is sufficient for evolvability. As we saw illustrated by Con-summarizes and extends the research he has been engaged in for several way's Game of Life, you have to have just the right sort of Order, with just decades, and lets us see for the first time how he himself places his ideas in the right mix of freedom and constraint, growth and decay, rigidity and the context of the history of the field.
fluidity, for good things to happen at all. You only get evolution, as the Santa Many have heralded him as a Darwin-slayer, finally driving that oppressive Fe motto proclaims, on the edge of chaos, in the regions of possible law that presence from the scene, and doing it, moreover, with the flashing blade of 222 BIOLOGY IS ENGINEERING
Stuart Kauffman as MetaEngineer
223
form the hybrid zone between stifling order and destructive chaos. Fortu-such landscapes, smooth or rugged, governs both the evolvability of pop-nately, our portion of the universe is poised in just such a zone, in which the ulations and the sustained fitness of their members. The structure of fitness conditions for evolvability are tuned just right. And where did those salu-landscapes inevitably imposes limitations on adaptive search. [Kauffman brious conditions come from? They could "in principle" have come from the 1993, p. 118.]
wisdom and foresight of a designer like Conway,
or
they could have come Notice that this is all pure Darwinism—every bit acceptable and nonrev-from a
prior
evolutionary process, either one with selection or one without.
olutionary, but with a major shift of emphasis to the role of the topology of In fact—and this, I think, is the heart of Kauffrnan's vision—evolvability the fitness landscape, which, Kauffman argues, has a profound effect on the itself not only must evolve (for us to be here), but is
likely
to evolve, is almost
rate
at which design innovations can be found, and the
order
in which design sure to evolve, because it is a forced move in the game of Design.20 Either you chances can accumulate. If you have ever tried to write a sonnet, you have find the path that leads to evolvability or you don't go anywhere, but finding confronted the basic design problem that Kauffrnan's models examine: the path to evolvability is not such a big deal; it's "obvious." The principles of
"epistasis," or the interactions between genes. As the budding poet soon design that make biological evolution possible will be found, again and again discovers, writing a sonnet isn't easy! Saying something meaningful—let and again, no matter how many times we rerun the tape. "Contrary to all our alone beautiful—within the rigid constraints of the sonnet form is a frus-expectations, the answer, I think, is that it may be surprisingly
easy"
trating exercise. No sooner do you tentatively fix one line than you have to (Kauffman 1993, p. xvi).
revise many of the other lines, and that forces you to abandon some hard-won When we considered forced moves in Design Space in chapter 6, we were excellences, and so forth, round and round in circles, searching for an overall thinking about features of the final products that were so obviously "right"
good fit—or, we might say, searching for overall good fitness. The that we would not be surprised to find them independently appearing—
mathematician Stanislaw Ulam saw that the constraints of poetry could be a arithmetic among the alien intelligences, eyes wherever there is locomotion source of creativity, not a hindrance. The idea may apply to the creativity of through a transparent medium. But what about features of the process of evolution, for just the same reason:
creating those products? If there are fundamental rules about how things have to be designed, about the order in which design innovations can be created, When I was a boy I felt that the role of rhyme in poetry was to compel one the strategies of design that are bound to work or fail, then these should be to find the unobvious because of the necessity of finding a word which homed in on by evolution just as surely as the features of the finished rhymes. This forces novel associations and almost guarantees deviations products. What Kauffman has discovered, I submit, is not so much
laws of
from routine chains or trains of thought. It becomes paradoxically a sort of
form
as
rules for designing:
the imperatives of meta-engineering. Kauffman automatic mechanism of originality. [Ulam 1976, p. 180.) has many telling observations to make about just such principles of meta-engineering that govern the process by which new designs could, in practice, Before Kauffman, biologists tended to ignore the prospect that evolution be created. We can consider them to be features of the whole phenomenon of would confront the same sort of pervasive interactions, because they had no evolution that have
already
been discovered, have
already
gone to fixation, clear way of studying it. His work shows that making a viable genome is in effect, in our part of the universe. (We will not be surprised to find them more like writing a good poem than simply jotting down a shopping list.
everywhere else in the universe where there are designed things, because this
Since
the structure of fitness landscapes is more important than we had is the only way to design things.)
thought (with our simpler, Mount Fuji models of hill-climbing), there are constraints on design-improvement
methods
that keep engineering projects Adaptive evolution is a search process—driven by mutation, recombina-channeled into more narrow paths to success than we had imagined.
tion, and selection—on fixed or deforming fitness landscapes. An adapting population flows over the landscape under these forces. The structure of Evolvability, the capacity to search a reasonable fraction of the space, may be optimized when landscape structure, mutation rate, and population size are adjusted so that populations just begin to 'melt' from local regions of the space. [Kauffman 1993, p. 95.]