Page 128 - Deep Learning

Creative Insight: The Redistribution Theory 111

A key feature of this explanation is that nothing extra or unusual is hap-
pening at the moment of insight. Both perseverance and insight are produced
by the same redistribution process. In every cycle of operation, some option
is tried, its outcome evaluated, and feedback is propagated down through the
system. In each choice point, negative feedback causes a certain amount of
activation to be subtracted from the unsuccessful option and redistributed
across all other options. This is what is happening when the person perseveres
with an unsuccessful approach, and also what is happening when a person has
an unexpected insight that leads to the solution. There is no difference in the
nature of the processing that occurs at those two qualitatively different types
of mental events. The difference between perseverance and insight is not due
to different cognitive processes but depends on the relative initial activation
levels, the threshold, the amounts of positive and negative feedback, the decay
rate and other factors. That is, the difference emerges out of the interactions
among the structure of the processing system, the current strengths and acti-
vation levels, the location of the threshold and the evaluations of the outcomes.
The punctuated nature of insight is not built into these processes, but emerges
in their interactions.
The computer model simulates the behavior of a single processing unit.
The hypothesis is that our perceptual systems comprise tens of thousands, per-
haps hundreds of thousands, of processing units, each of which functions in
accordance with the six redistribution principles. The behavior of the overall
system is a function of the structure of the network – which nodes are linked
and by which type of link – and the exact strengths and activation levels. In
some cases, subtracting and redistributing activation in one of the choice
points will have no effect. The option with the highest activation might still
be the most active option even after negative feedback has caused some of its
activation to be lost. If so, then the result is perseverance; that is, the prob-
lem solver keeps pursuing the same approach to the problem, although the
approach is known not to work. Perseverance is often observed in laboratory
studies of insight.
Another possibility is that the rank order of the activation levels of the
above-threshold options is revised when activation is subtracted and redis-
tributed. If the most-active option in the new rank ordering was above thresh-
old to begin with, no novel option comes to mind, but the person’s behavior
changes. He pursues another of the options that came to mind initially. If the
second option is also unsuccessful, control might pass to yet another of the
above-threshold options. Behaviorally, this appears to an observer as a deliber-
ate, heuristic search.