Page 143 - Deep Learning

126 Creativity

Wallas’s four-stage sequence is similar to the insight sequence in that it is
also a pattern of alterations in mode and tempo, but with the crucial difference
that an impasse is not an incubation period. During an impasse, the problem
solver is passive in the sense that he is not making progress, but this passivity is
enforced rather than voluntary; he does not know what to do next. During an
incubation period the problem solver decides to set the problem aside. While
it is true that people sometimes resolve impasses in the course of continued
attention to a problem, it can also be true that the probability of resolving an
impasse sometimes increases after a rest period. We lack data on which of
these two patterns is most prevalent but as long as both occur at least some-
times, both require explanation. But principles proposed to explain incubation
effects are not in competition with principles that explain the resolution of
impasses; the two sets of principles are complementary.
The literature on incubation is not extensive. C. A. Kaplan reviewed 18
63
studies, 12 of which provided positive evidence for incubation. The support is
not overwhelming but sufficient to make it more likely than not that a pause
can help. Why might it help to pause? It is paradoxical that doing nothing can
be more productive than doing something. One repeatedly rediscovered pos-
sibility is that the pause allows the problem solver to forget the unproductive
approach. In the words of R. W. Weisberg:

… forgetting is what is needed. A person may be unable to solve a prob-
lem because he or she is approaching it in the wrong way. If so, a break
might enable them to forget the incorrect approach, thereby making it
easier to think of a different, potentially more successful approach. 64

The problem with the forgetting explanation is that it does not explain why
the incorrect representation is not re-activated after the pause. If a problem
activates a certain representation when it is first encountered, then why would
it not activate that same representation when the problem solver returns to the
problem after the rest period? If the initial encounter with the problem is a race
between an unhelpful and a helpful representation, the forgetting that occurs
during the rest period will affect both representations equally. Forgetting, by
itself, does not explain how or why the balance between the two representation
shifts during the rest period. The forgetting hypothesis must be supplemented
with an auxiliary hypothesis about differential rates of forgetting to be a viable
explanation.
65
Simon proposed an explanation of this sort. He postulated that control
information, including subgoals, decays at a faster rate than factual informa-
tion about a problem. In addition, he assumed that factual information is only