Biology Reference
In-Depth Information
preventative changes in software and hardware, but also because many errors
were not fatal, and computer users were not unaccustomed to rebooting frozen
machines already. (Taking the larger system into account thus reduced the real
threat.) How about organic systems? For decades of progress in genetics, using
reductionistic analyses (in which more details more accurately known were
always better), we were presented with stories about how small changes in a
system may change its behavior radically - and indeed that was how genetic
changes were supposed to affect molar systems. The paradigm case for this view
was the now classic story in every textbook of how a single base substitution
in the gene coding for the beta-chain of the hemoglobin molecule could lead on
the one hand (in the heterozygote) to greater resistance to malaria, and on the
other to severe anemia and 'sickle-cell crisis' and an early painful death among
HbS homozygotes. And that was just the beginning: There was a long list of
'single gene' genetic diseases.
This presents a picture in which organisms are like computer programs, so
we need to know the genetic constitution and the biochemistry of the system
in great detail because a single change could wreak major havoc. Yet software
has an organization: We often do not need to know the details of a procedure to
write other parts of a program, and object-oriented programming has increased
the modularity of code substantially. While there are serious problems caused
by inaccurate data, the picture of organic systems that would require complete
knowledge to analyze any aspect of system behavior is not accurate: if it were
we would be unsurvivable, unevolvable, and unstudyable. In some ways, at a
very simple level we are like a house lighting system: There are things, like a
failure at the main junction box, that can shut down the whole thing, and blown
fuses can temporarily take out subsystems of varying sizes. But most failures
in the system are, and are designed to be, strictly local. Thus individual bulbs
and appliances can fail without requiring anything more than their replacement
or repair because of the parallel (redundant) organization of the house wiring at
the lowest level. Unreliable data are critical problems for the NSB, as they must
be for the analysis of any complex system,
8
but there are various mitigating
conditions.
Faced with the problem of unreliable data, we must find workarounds. Some
must come through improvements in technology and the development of better
means of testing the accuracy of our data. Some come through choice of questions
that are less affected by this problem, though this may skew research and theory
construction, as we saw earlier with the differences engendered by 'single' vs.
'multichannel' approaches, both for neurophysiology and for cell biology.
8
Taylor's (1985) study in ecological communities in nature and with simulations showed that interactions left
out could make studied components appear causally connected when they were not, or make them apparently
independent when they were not. The conclusions would apply for networks more generally.
