IDyOT: A Computational Theory of Creativity as Everyday Reasoning from Learned Information - Computational Creativity Research: Towards Creative Machines - page 128

be waking in the morning with the solution to a problem experienced the previous day
in mind. Deliberate creativity is a process experienced by the human conducting it 5
as a consciously controlled series of reasoning and development of a creative artefact;
an example would be the film composer who constructs a new theme tune to order
by the application of rules that she knows from experience will yield acceptable and
appropriate material. Many composers, among themMozart [ 21 ] and (less famously)
the current authors, view composition of music as a cyclic combination of these
things, where spontaneous ideation is followed by deliberated development, which
in turn leads to further spontaneous creativity, and so on.
Wallas [ 47 ] proposes a multi-phase process of creativity, in which “incubation”
leads to “illumination”; it seems likely that the “illumination” point, colloquially
known as “the 'Aha!' moment”, corresponds with the emergence of an idea fromnon-
conscious systems that emit it into conscious awareness. Indeed, to claim otherwise
necessarily leads into mysticism (which we eschew here) because there is nowhere
else for the ideas to come from.
IDyOT is conceived as a cognitive architecture, following Baars' Global
Workspace Theory [ 1 ], of which more detail is given below. A key detail of that
formulation is the throttling of a large amount of information produced in parallel,
by an explicit selection mechanism, as it approaches conscious awareness. The ulti-
mate effect of this mechanism is that an organism is conscious of one thing at a
time, and, conversely, consciousness, an expensive resourse, is managed and applied
only where it is required. Merker [ 30 ] and Shanahan [ 45 ] propose candidate neural
assemblies that may accommodate Baars' selection mechanism, but no progress has
yet been made towards empirical examination of these proposals as a mechanism for
creativity. This is partly because making testable predictions about creative behav-
iour is extremely difficult; it is to be hoped that simulations such as IDyOT may help
to do so in future.
For the purposes of the current work, it is not necessary to consider the nature of
consciousness (the “hard” problem [ 5 ]), but, instead, only what passes through the
throttle—a much easier problem; therefore, we refer to “conscious awareness”, to
make the distinction clear. Thus, the current, initial work on IDyOT is focused on
optimising the throttling mechanism, effectively a parameter of an abstracted parallel
sampling system, so as to allow the system to perform the various functions that we
predict should arise from the architecture. The original proposal for this parameter
is given by Wiggins [ 55 ], though we are considering a wider range of possibilities.
5 We avoid the troublesome question of whether the phenomenon experienced as a conscious
decision is really that, because it is not relevant here.