I thought it might be both useful to look briefly at several places
where science fails. In my attempts to understand phenomena and
fields, I always prefer a balanced approach, one that looks at the
strengths and weaknesses, at the virtues and the flaws.
1. Science makes progress through measurement, but not everything can
I once stated, "Science measures what it can measure and defines the
rest to be unimportant." Many others have said similar things, often
much more strongly. Einstein, for example said "Not everything that
can be counted counts, and not everything that counts can be counted."
There is an attempt to use precision, even if the thing that is being
measured is not really what is of concern. In Finance, for example,
the riskiness of a stock is specified as a single number, Beta, that
refers to the ratio of the variability of an asset to its portfolio
divided by the variability of the portfolio (Footnote 2). In fact,
this fails to capture what most of us would think of as risk. Risk to
most of us is not simply variability. But the financial community
likes it because it is a simple measure that is easy to compute and
makes the math work. See footnote 2.
In general, science has limitations on measurement, so it measures
what it can, often defining that to be the item of interest. In the
first few years of the method, it is roundly criticized, but as time
passes it becomes accepted and people no longer think to question it.
For numerous reasons, science dramatically oversimplifies the
phenomena under study. The major reasons have to do with measurement,
control, and mathematics.
A. Oversimplification caused by measurement
Science measures what it can, and quite often this is very restricted.
See Footnote 1. I see this in the practical problem of design. In the
development of intelligent automobile automation, the automation needs
to know how slippery the road is, so the engineers look at the
automobile wipers: if on, it must be raining. This is a 0th order
approximation to truth and dangerous, because roads are most slippery
when roads are damp, caused by light drizzle. Many drivers do not yet
use their wipers. (Most dangerous because light rain brings out the
oil in the road. heavy rain washes the oil away.)
It is easy to find situations where a simplified measurement is taken
to represent a complex situation. SOmetimes this is a good
approximation. Often it works well in the laboratory, but fails in the
B. Oversimplification caused by control.
It is very important to be able to control the phenomena under study.
This is why so many studies are done in the laboratory, where all the
variables defined to be extraneous and irrelevant can be controlled
(temperature, humidity, light, noise, ... ). In the human and social
sciences, this means putting people in extremely unnatural situations
and asking them unnatural questions, questions they would not normally
encounter, or even if they encountered them, they would be deeply
embedded in a rich context.
B. Oversimplification caused by mathematics
Mathematics is a powerful tool in enabling precise explanations and
predictions. But mathematics can get quite complex very rapidly. Until
recently, non-linear systems could not be handled, so everything was
linearized. Most important real phenomena are non-linear, which means
that much of theory was a simple approximation to reality over a
restricted range or, in the case of wildly non-linear phenomena, that
they couldn't be studied.
Economics is a good example of the dangers of oversimplification. The
fundamental assumptions in much of economic theory are logical and
sensible, but quite wrong as accurate descriptors of real human or
institutional behavior. Nonetheless, they greatly simplify the
calculations. As a result, much of the edifice of economics is built
on false premises. This fact has been known for decades (Herb Simon,
among others, made this point. But even when he got the Nobel for this
work, the economists ignored him. "What a waste of the prize," one
economist told me. I repeated the quote to Herb, who smile patiently.)
Now, with a second Nobel awarded to yet another truth-sayer (Danny
Kahneman) are mainstream economists starting to take this seriously.
(See item 4, below: Scientists are human.)
Today the maths are more powerful and computer simulation has added to
the power of formal models, but the need for simplification still
exists. But there is hope. The power of computers and the huge
databases that can now be assessed, plus the power of modern
visualization tools means that we can start looking at things a
complex, interdependent, systems (and dynamical systems), which is
much more appropriate for many of the real problems that we face.
These are still simplifications, but nonetheless, an improvement.
If you want to see some extreme simplifications in the field of
design, I point you to such works as Axiomatic Design. (Footnote 3)
3. Paradigms, Frameworks, and Fads
As philosophers of science like to point out (my favorite writer on
the topic is Bruno Latour), scientists are human (see point 4, below),
and subject to all sorts of fads. Each generation of science has some
accepted paradigm, and studies that go outside the framework of that
paradigm are often ignored, or rejected by the high prestige journals.
Fads come and go. Every so often, the young Turks come along and
introduce a new paradigm. This takes decades, and many die along the
way. Worse, once the new paradigm is accepted, it starts to explain
whole new phenomena (which is good), but often ignores a lot that had
been learned in previous paradigms (which is bad). I have seen
numerous such shifts.
I was a bystander watching this young kid Noam Chomsky overthrow
linguistics to the point where for a while much expertise in languages
was ignored in favor of theories of highly abstract and simplified
syntactical constructions (completely ignoring the difference between
written and spoken language). Molecular biology threw out the
systemic and general biologists who studied real animals and plants. I
was one of the young Turks who threw out behaviorism in psychology in
favor of, in temporal order, information processing, cognitive
psychology, cognitive science, cognitive neuroscience. Then came
connectionism (born in the office next door to mine) and dynamical
systems. I am now one of those old fogies whose work has been
overthrown by todays young Turks. (Which is as it should be.)
4. Scientists are human
Scientists are people, subject to the same failures as the rest of us,
so as they jockey for recognition and power, they tend to exaggerate
the importance of their findings and, much worse, conceal data, and
then even far worse, lie and fabricate. Scientists form cliques in the
honorary societies and in grant awards, where they reinforce their
colleagues and keep out their dissenters (or people who follow other,
opposing paradigms). Science is supposed to be open-minded. In fact,
science is one of the most conservative of fields. In many ways that
is good, because it does help prevent the rush to the latest fad. But
having a closed mind is never a good thing.
5. Although the scientific method means that science is
self-correcting, it may take decades, generations, or even centuries
for the process to play out for any given topic.
Yes, the scientific method is open and testable, which means that
eventually it converges upon an accurate description. But this process
might be too slow to help those who are caught in the pettiness.
1. In my Google search to get the Einstein quote right I came across a
wonderful article about the powers and dangers of measurement. It
applies very much to the debate on this discussion group:
"Measurement, a blessing and a curse."
2. Beta as a measure of risk. See the Wikipedia article, especially
the section called "Criticism":
3. An excellent introduction to and comparison of Axiomatic Design and
Triz can be found at
Nielsen Norman Group
[log in to unmask] www.jnd.org http://www.core77.com/blog/columns/
IDEO Fellow. Latest book: "Living with Complexity"