Page 130 - Deep Learning

Creative Insight: The Redistribution Theory 113

retrieval probes, nodes from which activation spreads through memory. As
a result, previously dormant but potentially useful knowledge items – cog-
nitive operations – are retrieved; they suddenly “come to mind,” as we say.
Newly retrieved actions, reasoning schemas and subgoals will in turn trigger
the retrieval of previously unheeded heuristics to control their application.
The upshot is that the heuristic search mechanism is operating within a dif-
ferent search space. The cognitive system has leaped sideways to override the
imperatives of its own prior knowledge. The new search space is also con-
strained, but constrained in a different way and hence might include a work-
able solution. In summary, unit alterations in problem perception caused
by the downward propagation of feedback cause changes in the problem
representation, which in turn causes the retrieval of previously unheeded
knowledge elements from long-term memory; the latter define a different
search space.
This hypothesis explains why problem solvers sometimes cease overt
activity, including concurrent verbalization, during an impasse. A person
trapped in an impasse is not in the same mental state as someone who has
disengaged from problem solving. The thinker is absorbed in the problem; he
is concentrating. His brain is busy propagating negative feedback down the
processing layers, recomputing the balance between options in individual pro-
cessing units and propagating the consequences back up to working memory,
where changes will trigger knowledge retrieval by altering the distribution of
activation over long-term memory, which in turn will trigger look-ahead in
the new search space. These processes conspire to create a false impression
of passivity, for two reasons. First, representational change processes do not
correspond to any particular overt behaviors. One does not spread activation
with a shovel. Second, representational processes tend to occur outside con-
sciousness. Immediately before an insight, the problem solver does not know
any better than an observer what is going on in his mind. He feels compelled
to concentrate on the problem, but he has little to say about what he is doing,
much to the chagrin of the psychologist who wants to know what happens as
an impasse is resolved.

Aftermath: Pursuing the New Option
What happens after the insight event? If breaking an impasse produces a
change in representation that triggers a change in what is retrieved from
memory, why does the problem solver sometimes feel as if the complete
solution suddenly came to mind at the moment the impasse was resolved?