Page 164 - Deep Learning

Creative Insight Writ Large 147

Creativity

Expertise

Low Middle High
Figure 5.4. The triangular correlation hypothesized to relate expertise and creativity.

levels of expertise, the person does not have enough background knowledge
to carry out the necessary analytical work, and hence is almost guaranteed
not to succeed. However, at high levels of expertise, the outcome is contin-
gent on the content and representation of that knowledge. If the inventor’s
understanding of his problem tends to activate helpful knowledge elements,
the probability of success is high; but if that knowledge is encoded in such a
way that the initial problem representation activates unhelpful knowledge ele-
ments, the probability of success decreases. This state of affairs should create a
particular empirical pattern, known as a triangular correlation. In research on
the related question of the connection between creativity and intelligence, the
expectation to find this type of pattern is sometimes referred to as the threshold
31
hypothesis. The key idea is that a person has to be above a minimal threshold
to able to succeed; but being above that threshold does not guarantee success;
it only makes it possible. This state of affairs implies that if we could measure
and plot expertise versus creative achievement for a large number of individu-
als, the result would resemble Figure 5.4. The evidence for a triangular correla-
tion between intelligence and creativity is weak and somewhat inconsistent;
evidence regarding the predicted triangular correlation between expertise and
creativity is not available. The important point for present purposes is that the
hypothesis that impasses are caused by the constraining effects of inappropri-
ately activated prior knowledge has a definite and testable prediction at the
significant time band.