Page 89 - Deep Learning

72 Creativity

including perception, decision making and problem solving. after the
32
appearance of his review, the application of Darwin’s theory of variation and
selection to cognitive phenomena became an academic growth industry. By
33
the end of the 20th century, the trickle of such applications had grown into a
flood that cannot be contained in a review.
The various articulations of the variation-selection principle exhibit less
fidelity to their source of inspiration than their advocates would have us believe.
One would expect a variation-selection theory of human creativity to specify
cognitive analogues to differential reproduction rates, environmental selection
among phenotypes, genes, genetic variation, ontogenesis, organisms, popula-
tions and species. as evolutionary biologists tirelessly point out, although the
organism is the unit of selection, the species (or the population) is what evolves,
so a Darwinian variation-selection theory must recognize at least three system
levels: gene, organism and species (or population). a key component of the
Darwinian mechanism is reproduction, the multiplication of those organisms
whose genes give them the advantage in the struggle for survival.
The various applications of the variation-selection idea to cognitive processes
do not fit this theoretical schema with any precision: What are the cognitive ana-
logues to the three system levels? What is the population that evolves in creative
thinking? What are the cognitive counterparts to ontogenesis and reproduction?
Critical analyses by David N. Perkins and others show that variation-selection
theories of cognition are not Darwinian in any precise sense, and so cannot draw
strength from the proven ability of biological evolution to produce novelty. 34
two key issues in formulating a variation-selection theory of cognition
are how the variations are generated and how the successive variation-selec-
tion cycles are related. With respect to the first issue, Campbell argued that
random generation is sufficient. as in generate-and-test theories, the notion
of a random generator frees the mind, and hence the theorist, from the need
for an informed generator.
anticipating resistance, Campell launched an energetic defense of ran-
dom generation as an explanation for creative achievements. First, the density
of useful novelties in a typical possibility space is indeed low, but we tend to
overestimate the success rate of attempts to create. if we take into account the
number of failures in creative fields like art, science and technology, the ratio of
success to failure will turn out to be consistent with a random search through
the space of possibilities, he claimed. But as long as no calculation or estimate
of this ratio is presented, this pseudo-quantitative argument remains a prom-
issory note. it also fails to explain why some individuals, studios, research and
development teams and laboratories consistently create more than others. On