CULTURE AND RESEARCH INFRASTRUCTURE 365
A related point is that computer scientists tend to assume that universal statements have no excep-
tions, whereas biologists have learned that there are almost always exceptions to rules. For example, if
a biologist says that all crows are black, and one asks about albino crows, the answer will be, “Oh, sure,
albino crows are white, but all normal crows are black.” The biologist is describing the average case—all
standard crows are black—but keeps in the back of his or her mind the exceptional cases. By contrast,
the computer scientist wants to know if the crow database he or she is building needs to accommodate
anything other than a black crow—and thus, when a computer scientist makes a biological generaliza-
tion, the biologist will often jump immediately to the exceptional case as a way of dismissing the
generalization.
These comments should not be taken to imply that biologists do not use theory at all; in fact,
biologists use theory and models in their everyday work. The theory of evolution is among the most
powerful of all scientific theories, in the sense that it underlies the scientific understanding of all natural
biological phenomena. But because the outcomes of evolutionary processes are driven by a myriad of
environmental and chance influences, it is difficult to make measurable or quantitative predictions
about specific biological phenomena. In this context, evolution is more of an organizing principle than
a predictive formalism.
Perhaps a fairer statement is that many biologists remain to be persuaded of the value of quantita-
tive theory and abstraction on a global basis, although they accept their value in the context of special-
ized hypothesis, individual probes, or inquiries on a biological process. Biological researchers are begin-
ning to see the potential explanatory value of computational and mathematical approaches—a potential
that is less apparent than might be expected because of the very success of an empirical approach to
biology that has been grounded in experiment and observation for many decades.
10.3.1.4 Data and Experimentation
As mentioned above, computer scientists and biologists also view data quite differently. For the
computer scientist, data usually result from measurements of some computational artifact in use (e.g.,
how long it takes for a program to run, how many errors a program has). Because these data are tied to
artifacts that have been made by human beings, they are as ephemeral and transient as the underlying
artifact, which may indeed change in the next revision or release. Because computer science is a science
of the artificial, the intellectual process of the computer scientist does not begin with data, but rather
with an act of artifact creation, after which measurements can be taken.^66
Indeed, for the computer scientist, the term “experimental computer science” refers to the engineer-
ing and creation of new or improved computational artifacts—hardware or software—as the central
objective of intellectual efforts.^67 Engineering has intellectual biases toward model reduction, extracting
key elements, and understanding subsystems in isolation before assembling larger structures. The
engineering approach also rests on the idea that basic units (e.g., transistors, silicon chips) have repeat-
able, predictable behavior; that “modules” with specific capability (e.g., switches, oscillators, and filters)
can be made from such units; and that larger systems with arbitrary complexity are, in turn, made of
such modules.
In contrast, biology today is a data-driven science—and theories and models are created to fit the
data. Data, presuming they are accurate, impose “hard” constraints on the biologist in much the same
way that results from theoretical computer science impose hard constraints on the computer scientist.
Because of the central role that data play in biology, biologists pay a great deal of attention to experi-
(^66) This is not to deny that computer scientists often work with large datasets. For example, computer scientists may work with
terabytes of textual or image data. But these data are the subjects of manipulation and processing, rather than being tied directly
to the performance of the hardware and software artifacts of the computer scientist.
(^67) National Research Council, Academic Careers for Experimental Computer Scientists and Engineers, National Academy Press,
Washington, DC, 1994.