the book: The Knowledge Machine by
Michael Strevens reviewed by: Stephen P.
Cook date of review:
April 2021
Science can be defined as a body of knowledge ultimately based on observation obtained by
application of the scientific method. If forced to provide just
a three word definition, one could do worse than using The
Knowledge Machine. This is the title of a recent book by NYU
Philosophy Professor Michael Strevens. In doing this, “machine”
stands in for “the scientific method.” Part I and this four part book is titled
“The Great Method Debate.” It presents nice account of opposing
view of how science works—and in particular contrasting the views of
Karl Popper and Thomas Kuhn. I generally have faith in science’s
built-in, self-correcting mechanism that depends on healthy skepticism
to weed out what is not true. When properly functioning, I believe
erroneous judgments will be rare, and when they are made they will
soon be fixed. Strevens has much less faith.
Granted humans are imperfect creatures and they can use science in
perverted ways. As Strevans reports, CUNY sociologist Stanley
Aronowitz has written, “Science legitimates itself by linking its
discoveries with power, a connection which determines what counts as
reliable knowledge…The strongest team decrees what counts as
truth.” I disagree with such suggestions that scientific knowledge
is subjectively arrived at. And
with those like Strevens
who attack science by saying things like, “the logic of
scientific reasoning is by its very nature subjective.”
Strevens’ book nicely details how modern science developed only
after the need to compartmentalize and avoid “subjective
considerations and non empirical considerations (philosophical,
religious, aesthetic) from official scientific argument” was
appreciated. And that, while a great strength of science is an
open-minded curiosity that asks questions and make connections through
a wide swath of space and time, science also depends on minds narrowly
focused. Not close-minded, but at times necessarily focused to
“carry out tedious measurements and perform costly and time
consuming experiments” as he describes it.
While Strevens’ book illustrates how human knowledge advanced
in great strives once people realized the need for evidence gathered
by careful observation and empirical testing, he paradoxically
subtitled it, “How
Irrationality Created Modern Science.” I fear that will attract anti-science
readers, who’ll get
encouragement from claims like “…in
their thinking about the connection between theory and data,
scientists seem scarcely to follow any rules at all.”
Were the author of this ridiculous statement here, I would patiently
pull several books out of my library to convince him of his folly.
I’d begin with Bright’s book. I’d shove his nose in the
mathematical details of the 150 pages spanning four chapters with
titles, ”Classification, Sampling, and Measurement,” “The
Analysis of Experimental Data,” “Errors of Measurement,” and
“Probability, Randomness, and Logic.” I’d point out that
scientists were trained with books like this seventy years ago.
And that today—with improved techniques for statistical
analysis of data—they’re undoubtedly getting even better equipped
to not do what Strevens implies is often done: getting away
with cheating. Strevens’ 350 page book has exactly three paragraphs
related to statistical analysis. Such techniques, he says “can be
gamed to illuminate the data from the most favorable (or publishable)
angle.” I say he’s unfairly connecting scientists with a
professional ethics
failing.
Despite having some educational value, I think Strevens’ book will
eventually be placed in the category that philosopher Thomas Kuhn’s
1962 book The Structure of Scientific Revolutions has been
placed in by many: contributing to “the debasement of science and
the debasement of truth.” Those are words of film-maker and onetime
graduate student of Kuhn’s Errol Morris. Rather than developing
through a gradual, cumulative process, Kuhn emphasized the importance
of revolutionary periods of paradigm shifts. He argued that those on
opposite sides of one of these shifts can’t communicate because of
fundamental differences in worldviews. And that with those differences
come different ways of doing science. Critics say he introduced a
subjectivity, cultural relativism, and an irrationality into how many
viewed science. Alexander Bird describes Paul Hoyningen-Huene’s
understanding of Kuhn’s position as follows. “We cannot possibly
find out whether a theory is true for that requires that we are able
to compare the theory and reality, which in turn requires having an
independent grasp on what reality is like. And that is precisely what
we do not have—and if we did have it, we would not need the
theory…” Morris sees a line from Kuhn to “alternative facts”
famous Kelly Ann Conway to truth-challenged Donald Trump.
While Strevens is seemingly not a radical subjectivist, at times he
appears sympathetic to their argument. I’m especially disgusted with
his treatment of the data that emerged from the 1919 solar eclipse
expedition effort to test Einstein’s theory. He became
famous after a team left by Sir Arthur Eddington essentially confirmed
one of his general relativity-based theory predictions. Einstein had
predicted that starlight passing near the Sun would be bent a slightly
greater amount than Newton’s theory predicted. Michael Strevens’
book spends several pages taking Eddington to task for subjectivity in
analyzing the data. He even reproduces a table from the expedition’s
scientific report showing deflection amounts for 18 stars and blasts
Eddington for throwing it out when it didn’t give the result
Eddington supposedly wanted.
Non-scientists often don’t
understand that all measurements have some associated error or
uncertainty. I took one look at the data for those 18 stars, put it in
a spreadsheet, and calculated a standard deviation measure of that
uncertainty. And I saw why Eddington justifiably threw it out that
associated error or uncertainty was very large. The average deflection
was 0.86 arcseconds— but the uncertainty was plus or minus 0.47.
Streven says nothing about uncertainties associated with the
measurements
Eddington decided the poor data was a result of the telescope mirror
getting too hot and expanding blurring the star images on the plates,
making them difficult to measure. As someone who years ago spent hours
measuring star positions on such plates—something philosopher Strevens
has probably not done— I
can certainly appreciate Eddington’s decision to throw out the data.
Likewise I doubt Strevens regularly looks at data,
calculates standard
deviations, and appreciates the important of those uncertainties.
In contrast, physicist Clifford Will—in his book Was Einstein
Right? —is careful to include that critical information. He
reports the two independently obtained data sets Eddington used as
follows: one based on eight photographic
plates gave deflection value of 1.98 arcseconds with plus or minus
0.12 uncertainty; the other based on just two plates gave 1.61
arcseconds with plus or minus 0.31. The uncertainties are much
smaller, indicative of much better data. Averaging the two values
gives a result rather close to Einstein’s prediction of 1.75
arcseconds deflection. Does it conclusively prove Einstein was right?
No. Strevens does provide an account of Karl Popper,
famed philosophy of science guy who later took a dim view of Thomas
Kuhn, being impressed with Einstein’s “willingness to subject his
theory to empirical tests that might disprove it.” As Popper put it,
“tests which could refute the theory tested, never establish
it.”
Elsewhere in his book Strevens, in one sentence, goes
after 1923 Physics Nobel Prize winner Robert Millikan for omitting
“many measurements that did not ‘look right’’’ What is being
questioned is data from Millikan’s famous oil drop experiment, from
which he derived the charge on the electron. I’ve spent hours with
students in physics labs doing a version of Millikan’s
experiment—and watching a Caltech produced video account of
Millikan’s data handling. And a lifetime dealing with problematic
data gathered by instrumental setups either compromised in some way,
or operating with low signal to noise rations. I again have no problem
with what Millikan did. I won’t completely diss Strevens’ book
because I learned some things from it, and it got me thinking. But,
after overcoming my instant dislike of its subtitle and reading it
anyway, I was disappointed with where it went.
Here’s another review: The
Knowledge Machine by Michael Strevens (book review by Jennifer Szalai
in The NY Times Oct 7 2020)
attention authors, readers, movie goers: If you have a book or film
whose themes you think are especially relevant to Project Worldview and
would like us to consider reviewing it please contact us as
follows:
Your
Suggestion: email it to us, put
suggested book (or film) review in subject line and click
here
|