Page 124 - Deep Learning

Creative Insight: The Redistribution Theory 107

principle, every person has a unique life trajectory and hence a unique knowl-
edge network, but subjects in experiments on insight – mostly college students
in Western universities – share enough cultural background that there are in
practice enough similarities across individuals for some problems to reliably
trigger impasses. This is the source of the persistent but erroneous intuition
that some problems are insight problems in some intrinsic sense.

How Impasses Are Resolved

The cause of unwarranted impasses is half the explanation for the insight
sequence. The other half is the mechanism by which impasses are resolved. If the
initial representation of the problem constrains the search space in unproduc-
tive ways by not retrieving the most useful knowledge elements, how can this
state of affairs ever change? The problem solver is trapped in a circle: Because
he thinks about the problem in a particular way, he retrieves certain concepts,
schemas, strategies and so on. Because he retrieves those particular knowledge
elements, he thinks about the problem in the particular way that is consistent
with them. To explain how impasses are resolved is to specify a mechanism
that enables the mind to break out of this unproductive circle.
Only change begets change, so the problem appears intractable: If we
attribute an insight to some cause C, then we have merely moved our focus
one step backward in the causal chain. To complete the explanation, we have
to explain why C happened when it did rather than earlier or later; this is pre-
sumably due to some prior cause Cʹ; and so on. All explanations of insight that
postulate special insight processes are undermined by this regress. A genuine
explanation has to terminate the regress, preferably by showing that impasse
and insight are two sides of the same coin.
Impasses cannot be resolved by pushing forward, searching deeper in the
inappropriately constrained space. Instead, the problem solver must draw back
to the initial problem representation to leap in a new direction. In the words
of Newell, Simon and J. C. Shaw: “What is needed in these cases [that require
an unconventional response] is not an elimination of the selective power of a
solution generator, but the replacement of the inappropriate generator by an
appropriate one.” 43
The triggering factor for this change is the problem-solving effort itself.
Persistent but unsuccessful solution attempts cause negative feedback to be
passed back down the layers of processing units. The experience of failure –
more generally, a negative evaluation of the outcome of a problem-solving
step – causes activation to be subtracted from the processing units that were