Page 134 - Deep Learning

Creative Insight: The Redistribution Theory 117

representations, and representations have parts. They are structured combina-
tions of simpler representations, and they can be revised by reordering and
replacing their parts. A representation is an interpretation, not a recording
of objective reality, so a problem does not determine its own representation;
alternative interpretations are always possible. A system that is able to repre-
sent is thereby also able to represent differently.
Second, what are the key features that distinguish creative processes
and justify calling them creative? In analytical thinking, the initial solution
space contains the desired solution, so the latter can be found by heuris-
tic search through that space. In creative processes, the analytical process-
ing is punctuated by one or more representational changes that revise the
search space. Creativity is a categorical dimension; any one thought process
either contains at least one non-monotonic representational change or it
does not.
Third, what gives direction to the creative process? According to the
present theory, the main determinant is the structure of the person’s process-
ing network. The amount of negative feedback and how it is propagated down-
ward through the processing layers will determine which links are suppressed,
and the content of the network will determine which options come to mind
instead. The direction of an insight event is massively contingent and for all
practical purposes unpredictable.
Fourth, what are the limiting factors? Why is it difficult to create? Our
brains are wired to constrain the options we consider by projecting prior
experience onto the present. Due to our evolutionary history, our disposition
to push forward, to apply prior knowledge more energetically or with more
care, is stronger than our disposition to draw back to leap. This bias is effec-
tive within tight contexts – that is, in stable, local situations of short duration.
For the mind to override its own bias, it must feed information back down the
processing layers to tip the balance among the options at some choice point
or another. But the relevant information might not be available. Not all task
environments are transparent with respect to the effects of actions. Negative
feedback might also be lacking because the motivation for persistence in the
face of failure is not present. Even if the person persists, the feedback might
not be strong enough to overcome the built-in bias to push forward instead of
drawing back to leap. Finally, no amount of feedback will help, if the person’s
mind is not prepared. It must contain the knowledge needed to construct the
alternative representation. Representational change can only resolve unwar-
ranted impasses. The redistribution theory thus answers the four creativity
questions.