Page 122 - Deep Learning

Creative Insight: The Redistribution Theory 105

A less Goal
constrained state
initial state

Dead-end
states

Initial
state
Initial, overly
constrained
search space
Figure 4.3. The relation between insight and search: Insight requires drawing back to
an initial state that generates a less constrained search space.

assumptions which restrict his sector of exploratory activity … and as long
as he continues actively at work in this section he does not escape from these
assumptions.” 39
Problems that end up on the psychologist’s list of insight problems have
a high probability of triggering an initial representation that in turn has a
low probability of activating those knowledge elements that are needed for
the solution, even when those elements are present in memory. A problem
becomes an insight problem for a person if it is misleading. The person’s prior
knowledge accidentally matches the features of the problem so as to suggest
that the problem should be understood in one way, whereas it would be more
productive to think of it some other way. The resulting solution space is overly
constrained and excludes the desired solution; see Figure 4.3.
Constraining contacts between prior experience and particular visual or
verbal problem features are indeed common in the classical insight problems.
40
Consider once again the Hat Rack Problem: Construct a hat rack out of 2-by-4
wooden boards and a C-clamp. The wedge solution to this problem consists of
two boards that are wedged between floor and ceiling and held in place with
the clamp; the hat is hung on the handle of the clamp. Why might this solution
be hard to think of? Everyday experience provides few examples of construc-
tions that rely on the ceiling as a source of support, but many examples where
2-by-4s are used to prop something up. This leads most people to attempt a
stable construction that stands on the floor. Visual attention is rarely directed