Page 150 - Deep Learning

Creative Insight Writ Large 133

not to choose between them but to discern which flavor best explains any
one phenomenon. A complex system is likely to be characterized by multiple
potential scaling variables and we cannot know before we conduct the rele-
vant analysis which scaling flavor is operating for a particular system and a
particular variable.
Regardless of which flavor applies in a particular case, there is no reason
to expect processes at level N to explain all patterns at N+1. Novel phenomena
and patterns are likely to emerge at the higher levels, limiting the explana-
tory power of events at the lower levels and requiring explanation in terms
of processes and mechanisms that are unique to the higher levels. Applied to
the production of novelty, these observations generate three questions: Which
characteristics of insight project onto higher levels of scale? What are the
mechanisms of projection? Which novel characteristics and mechanisms
emerge at the higher levels?
The production of novelty in significant projects differs from problem
solutions observed in laboratory studies along at least three dimensions: the
complexity of the tasks undertaken, the time required to complete them
and the number of people engaged. For lack of a better term, I will refer to
the third dimension as collectivity. The upper levels of time and collectivity
merge in history, which in turn is a single event on the evolutionary time
scale.

SCALING ACROSS TIME AND COMPLEXITY
There is no widely accepted metric of cognitive complexity and no natu-
ral levels of complexity present themselves to intuition. Was the invention
of radar more or less complex than the discovery of the structure of DNA?
Perhaps cognitive psychologists will one day specify system levels that are
as distinct as the cell-organism-species levels in biology or the atom-mole-
cule-substance levels in chemistry, but they have not yet done so. However,
by whatever measure, insight puzzles are less complex than significant cre-
ative projects. If so, to what extent are the principles introduced in Chapter
4 helpful in understanding the unfolding of creative projects? Do the pos-
tulated processes scale up from laboratory tasks with few components and
short solutions to scientific discoveries, technological inventions and works
of art? Because the theory of insight is built on top of the theory of analytical
problem solving, the natural approach is to verify that the latter is relevant,
document the occurrence of impasses and insights and then inquire into the
mechanisms of scaling.