Page 136 - Deep Learning

Creative Insight: The Redistribution Theory 119

succeeding a few minutes later is a case in point. This phenomenon has been
documented in laboratory studies under the label the tip-of-the-tongue phenom-
50
enon. In studies of list learning, Endel Tulving and co-workers noticed that
successive attempts to recall a list, first without and then with retrieval cues,
51
often resulted in better recall on the second, cued, attempt. If an item on the
list is recalled in attempt N+1 it must be encoded in memory, so if it was not
recalled on some previous attempt N, then the retrieval process failed to access it
and deliver it to working memory on that occasion, even though the item was in
fact available in memory. Failure to retrieve previously stored information in the
course of problem solving is also commonly observed in educational research,
where it is called inert knowledge. The existence of retrieval failure is so uncon-
troversial as to be a standard topic in textbooks in cognitive psychology.
The hypothesis that cognitive structures are associated with different lev-
els of activation is also a standard component of theories of human cognition.
It plays a central role in theories of memory, decision making and perception.
That working memory, and hence mental look-ahead, is limited with respect
to capacity is perhaps the most basic of all cognitive principles. The psycholog-
ical reality of mental look-ahead and heuristic search is supported by think-
aloud studies of problem solving. Finally, the notion of a threshold is neither
exotic nor ad hoc. Neural matter exhibits such a threshold: Individual neu-
rons respond to incoming stimulation by passing on the signal, but only if the
incoming stimulation is strong enough. The relation between a single neuron
and a cognitive processing unit is at the present time unclear, but it is plausible
that a processing unit that is built out of elements that operate with a thresh-
old will itself exhibit threshold behavior. In short, the redistribution theory of
insight relies entirely on concepts and principles that have independent sup-
port in cognitive research.
The theory explains insight as a side effect of the same processes that are
involved in analytical thinking. The cognitive processes that are executing in
the mind of the person who experiences insight are the same as those running
in the mind of the same person a minute earlier, or of another person who is
unable to think of a new idea. At each moment, feedback is propagated through
the system and activation is redistributed over all available options in each rel-
evant processing unit. This point cannot be stated strongly enough: According
to the redistribution theory, the experience of creative insight is a consequence of
the same cognitive processes as the experience of being at an impasse; the pro-
cesses that generate a novel approach are the same processes that produce per-
severance in an unsuccessful approach. No additional process is assumed to be
executing at the moment of insight over and above those that execute at any