Biology Reference
In-Depth Information
something we would have a right to expect were we not talking about evolved
systems, but it is bound to help with the analysis of complex systems in the
presence of unreliable data.
6. GENERATIVE ENTRENCHMENT
Robustness is interesting not just because it makes organisms survivable and
evolvable, but because robustness itself seems to be so pervasive among organ-
isms - in a word, so robust.
16
This is surely why there is so much interest in
studying the formal properties of networks, and also why the NSB should not
see itself in contrast with evolutionary biology, particularly the new evolutionary
developmental biology. Are there other features of the internal complexity of
the organism which have this kind of generality? There is one that has signifi-
cant implications for the behavior of networks: The architecture of development,
prima facie, can be used to predict differential rates of evolutionary change in
different factors, and identify constraints on how and how much they can change
(though generally not the details of how they will change). This is generative
entrenchment, and in particular, the differential generative entrenchment of dif-
ferent elements in a causal network (Wimsatt, 1986, 2001; Schank & Wimsatt,
1988, 2000; Wimsatt & Schank, 1988, 2004).
Differential entrenchment is not an accidental feature of evolutionary sys-
tems. It is generic. Nor is it avoidable in any of our engineered systems. The
importance of generative entrenchment points naturally to a number of architec-
tural and dynamical network properties, particularly redundancy and canalization
(ways of getting robustness) and modularity. Each of these act to modulate
and commonly to reduce its effects and magnitude. These should all qualify as
general properties of interest to the NSB, but they are also of central interest
to developmental genetics and evolutionary developmental biology. By collab-
orating in their analysis, the NSB extends its central importance to these other
disciplines.
If we consider a network, a pathway, or a cascade whether of gene activity
or of biochemical metabolism, different nodes are differently connected. If we
draw a directed graph for the propagation of causal effects in one of these or
in any mechanism - including any of the engineered products of our modern
technology - we will find that different numbers of nodes are reachable from
different starting points in the network. Figure 1 is a randomly constructed
directed graph of 20 nodes with 20 edges, generated by computer for our first
16
This is at least partially due to reasons suggested by Aldana and Cluzel (2003), but Wagner (2005) also
provides multiple arguments to this cumulative conclusion.
