Page 174 - Deep Learning

Creative Insight Writ Large 157

College, London, there was the research group led by Linus Pauling at the
California Institute of Technology. Each team explored different assumptions
about the molecule’s structure. For example, it was not known with certainty
how many backbones, strands that held the molecule together lengthwise, the
DNA molecule had. Nor was it known how the backbones were arranged. Did
they form a central core, as the label “backbone” suggests, with the informa-
tion-bearing elements attached to it (the “outside” solution), or did they form
a shell, with the information-bearing elements on the inside? Wilkins pro-
posed a single-strand model, Pauling a three-strand model, while Watson and
Crick began with an alternative three-strand model but soon homed in on
the two-strand, outside model that turned out to be correct. The possibility
space may seem small (2 3 = 6 possibilities), but evaluating any one of them
*
required much empirical and theoretical work. The relative speed with which
this fundamental scientific problem was solved was in part due to the parallel
explorations of that space. It bears repeating that this effect is not due to the
division of labor as ordinarily understood (different teams attacking different
components of an overall problem), but to multiple teams exploring different
solutions to one and the same problem simultaneously.
The micro-theory of insight claims that an insight can appear only if there
is relevant feedback, more commonly called criticism when the topic is collective
problem solving. It follows that groups that allow and tolerate mutual criticism
will be more creative than groups in which criticism is rare or considered inap-
propriate. This accords with the common observation that open sharing of infor-
mation and the practice of subjecting claims and proposals to public scrutiny are
key determinants of scientific progress. Unlike the prediction about the power of
51
diversity, the prediction about criticism is well supported by available data. Kevin
Dunbar and co-workers have conducted field studies in chemistry laboratories
that pertain to this point. They found that critical comments were common in
laboratory meetings, and that the most common response is methodological: An
unexpected finding is initially attributed to an error or to some undesirable or
previously unattended feature of the laboratory procedure. However, when unex-
pected effects persist, researchers seek a substantive, theoretical explanation that
can involve a revision of the relevant hypothesis. The culture of science does not
allow publicly stated criticisms to be ignored; they have to be answered. From this
point of view, communication is a powerful source of feedback.
Like individual creators, creative collectives are sensitive to externali-
ties. A well-documented example occurred in the discovery of the structure
of DNA. Watson and Crick were at an impasse in part because they worked
with the wrong structural forms of the molecules that form the horizontal