Page 95 - Deep Learning

78 Creativity

Most important, it specifies with precision the interaction between a general
cognitive mechanism (the choose-act-evaluate structure of the search process)
and task-specific knowledge (strategies and evaluation functions). The func-
tion of the latter is to tame the combinatorial explosion and provide selectivity,
either by reducing the number of options or by enabling evaluation of out-
comes. Nevertheless, the heuristic search theory, as applied to creativity, shares
weaknesses with other articulations of the accumulation principle.

evaluation of accumulation Theories

accumulation theories in general and the heuristic search theory in particular
explain how novelty is possible by showing how it can arise out of a succes-
sion of steps: if elementary actions are executed in a novel sequence, their
ultimate result might be novel. The crucial advances over generate-and-test is
that evaluation happens continuously, to each individual step along the way,
so each generate-and-test cycle is informed by, and builds on, the results of
prior cycles; see table 3.2 for a comparison of the essential features of some
prominent accumulation and variation-selection theories.
But accumulation theories cannot explain what is creative about creative
thinking. if every thought process is a search through some problem space, then

Table 3.2. Four theories of creativity based on variation and selection.

Theoretician/year Unit of variation Source of variability Basis for
and selection selection
e. Thorndike 1898 Physical actions Unspecified; iteration aftereffects;
through the behavioral satisfiers and
repertoire (?) annoyers
D. t. Campbell any cognitive random choice Performance
1960 process; outcomes
memory,
perception,
thinking,
deciding, etc.
a. Newell and Steps during Search heuristics; evaluation
H. a. Simon problem solving, task-specific functions
1972 mental or knowledge
physical
W. G. vincent Design features Wish to improve; results Flying charac-
1993 of experiments teristics