Page 331 - Deep Learning

314 Conversion

Falsification explains why scientists abandon an established but incorrect
theory, but Popper famously declined to speculate on the origin of new
theories. But a theory change is typically a change from one theory to another,
50
so without an account of the origin of the alternative theory, an explanation for
scientific progress is seriously incomplete. Furthermore, it would be peculiar
if the right epistemology of science turns out to be at variance with scientific
practice. in Popper’s description, scientists appear as logic machines, ready
to abandon a well-established theory at the drop of a fact. Historians of sci-
ence unanimously agree that this is not an accurate description of scientific
practice. researchers in the successful sciences do not strive to disprove their
favorite theories. They sometimes test theoretical predictions, but they are
overjoyed when those predictions fit their data, however fallacious this reac-
tion might seem to a logician. The ratio of published research articles that
claim support for the author’s theory to those that report falsification must be
99 to 1 or larger. in 35 years of reading psychological research articles, i have
never encountered the latter type of article by an experimental psychologist.
of course, scientists happily publish articles that falsify somebody else’s theory,
a practice better explained by the theory of resistance than by the principle of
falsification.
At an abstract level, Popper was of course right. scientists abandon theo-
ries because they are found to be false. But falsification belongs to the sphere
51
of logical forms; it should not be interpreted as a statement about what scien-
tists do as they go about their daily work and even less as a description of their
cognitive processes. in A. newell’s terminology, Popper’s account of theory
change was a knowledge­level account; that is, it abstracts from the relevant
processes and describes their outcomes in terms of knowledge states. in this
52
interpretation, Popper’s falsification principle states that if a scientist knows
that his theory implies a particular consequence and if he also knows that
the consequence is contrary to fact, then he knows (in an abstract sense) that
his theory is false. But this knowledge-level description does not specify how,
by which processes, a scientist concludes that a theory is false nor how it is
replaced.

The Accumulation of Anomalies
The philosophy of science matured into a distinct subfield of philosophy at the
time of the post–World War ii expansion of Western university education, and
the new field proved popular with students. The number of articles and books
about theory change grew rapidly. But the positivist attempts to formalize the