Page 107 - Deep Learning

90 Creativity

lack relevance for the study of creativity. The set of natural tasks would be
reduced to eating, gathering berries, procreating, sleeping and escaping from
cold winds by huddling in a crevice – an unpromising set of activities to study
when searching for a theory of creativity. If the criticism of artificiality is not
to be reduced to this absurdity, those who issue the charge of artificiality need
to explain how “artificial” is to be defined instead. Until they do, the charge of
artificiality has no meaning, and the artificial versus natural distinction is not
useful for judging the relevance of particular tasks for the study of creativity.
There are nevertheless difficulties with the idea of labeling problems as
insight problems. The first is empirical. For some of these problems, empiri-
cal studies support the claim that they engage other cognitive processes than
problems that are not designated as insight problems. However, for other
10
problems on the received list, there exists no empirical demonstration that
they do, in fact, elicit insights in experimental subjects. Many are designated as
insight problems on no other basis than that they seem to be insight problems
to the researchers who use them. As research continues, the list of problems
anointed as insight problems grows by an undisciplined mixture of received
but unexamined opinion, implicit appeals to readers’ intuitions and willing-
ness to mindlessly follow prior practice.
A second difficulty cuts deeper. Labeling a problem as an insight problem
is questionable because a problem is unlikely to elicit insights in everybody
who tries to solve it. Being an insight problem is not an objective property of
a problem in the same sense in which having a certain weight is an objective
property of a bowling ball. Insight is an event that arises in the interaction
between a person and a problem, and that interaction is determined by the
person as much as by the problem. Individuals differ with respect to which
ideas and concepts they have become familiar with throughout prior learning,
so even though one person needs a new idea before he can see the solution
to a certain problem, somebody else might be able to solve the same problem
through analytical thinking. For example, the Two-String Problem might not
be an insight problem to someone who works with pendulums, because he
thinks pendulum as soon as he sees a hanging rope. For him, the problem does
not require an act of creation and the problem is not an insight problem; for
others, it is.
To label particular problems as insight problems creates the temptation,
widely succumbed to in an implicit way, to assume that every time someone
attempts to solve one of the problems on the received list, his performance is
an instance of something called “insight problem solving.” This label, in turn,
suggests that all such solutions are of a kind, and therefore should be explained