Page 428 - Deep Learning

Notes to Pages 59–65 411

Berlin Structure-of-Intellect Test and four measures of creativity to range from
.36 to .54 in a sample of 1,328 German students; when corrected for processing
speed, the range was .07 to .24 (p. 164). In a meta-analysis of 447 correlations,
Kim (2005, p. 59), found that although the range of values varied between –.41
and .71, the mean correlation was .17 and only a handful of correlations was
greater than .41. A correlation of .40 means that 16% of the variation in creativ-
ity is accounted for by variation in intelligence, leaving 84% of the variation to
be explained by something else. In short, correlational studies show that indi-
vidual differences in creativity are not explained by individual differences in IQ.
The situation is the same for children; see, e.g., Fuchs-Beauchamp, Karnes and
Johnson (1993).
18. Harvard scholar David Perkins has developed the interesting notion of dispo-
sitional intelligence (Perkins et al., 2000; Tishman, Jay & Perkins, 1993). People
possess approximately the same cognitive capabilities and capacities, but they
are differently disposed with respect to their use. People whom we tend to regard
as exhibiting high cognitive ability or unusually high creativity are disposed to
engage those capabilities at the drop of a hat. For example, when confronted
with a particular set of facts, some people immediately seek to understand their
implications. The conclusions they come up with might not require extreme
cognitive capacity or complex processing that is beyond the reach of the rest
of us. It is the fact that they are disposed to seek such implications that distin-
guishes them. Applied to the production of novelty, this concept implies that
creative individuals are highly disposed to, for example, look for novel ways of
doing something.
19. For lifetime creativity studies using a cognitively informed, biography-like
approach, see, e.g., Gardner (1993) on Freud and Einstein, Gruber (1974) on
Darwin and others. Gruber and Wallace (1999) and Howe (1996) describe the
methodology of using case studies in this way. This approach differs from the
quantitative approach to lifetime creativity used by Galenson (2006), Simonton
(1988, 2004) and others.
20. See Metcalfe (1986) and Metcalfe and Wiebe (1987). Insight problems can also
be differentiated from noninsight problems with correlational methodology
(Gilhooly & Murphy, 2005).
21. See Bowden and Jung-Beeman (2003), Bowden, Jung-Beeman, Fleck and Kounios
(2005), Jung-Beeman et al. (2004), Kounios et al. (2006), Luo and Knoblich
(2007) and Luo, Niki and Knoblich (2006).
22. For example, see Kershaw and Ohlsson (2004) and Knoblich, Ohlsson, Haider
and Rhenius (1999).
23. To choose 5 objects among 10, pick any 1 among the 10, any 1 among the remain-
ing 9, any 1 among the remaining 8, etc., for a total of 10*9*8*7*6 = 30,240 dif-
ferent selections. Suppose that the objects can be related to each other via 10
different binary relations, i.e., relations that can be absent or take either one of
10
two values. This provides 3 = 59,049 different relational structures for each
selection of 5 objects. Multiplying these numbers gives us 1,785,641,760 possible
configurations.
24. See Miller (1996) for the estimate of vocabulary size.