Page 161 - Deep Learning

144 Creativity

(a) Time scale t
Rate of Insight
progress Search
+ Impasse

M
– Time

0 10 90 100

(b) Time scale 10t
Rate of
progress
+

M
– Time

0 50 100
Figure 5.2. Rate of progress when insight sequences are evenly distributed over time.
Panel (b) shows the same process as Panel (a), but at an order of magnitude lower
temporal resolution.

time scale as a period when progress was more rapid than average. Figure 5.2
graphically illustrates the (unlikely) case of even distribution of impasses with
respect to onset, severity and resolution. Panel (a) shows the process at a level
of temporal resolution at which the variations in the rate of progress over time
are visible. In panel (b), which depicts the same process at a lower temporal
resolution, the evenly spaced variations disappear from view and the process
looks smooth, as if it were progressing at a constant pace.
Figure 5.3 exhibits a case in which the severity of impasses varies over
time. For simplicity of exposition, it only covers two cases: a series of impasses
of low severity, followed by a series of impasses of high severity; see panel (a).
As in Figure 5.2, panel (b) shows the same process at a temporal resolution that
is an order of magnitude lower. Once again, the local variations in tempo fade
from view, but the average differences in rate of progress over a period of time
stand out. A casual observer might conclude that the initial 50 time units con-
stitute “a creative period” or an “annus mirabilis,” while the last 50 time units
constitute a “period of stagnation” during which something interfered with