Information Technology Reference
In-Depth Information
model can also be applied in these cases. A sketch can still be regarded as a material
form in relation to a more abstract conceptual representation. It is the difference in
level between the two representations that is important, and the interplay between
them when going back and forth—not the exact nature of the material representa-
tion. In the case of score-based music, for example, the material representation (the
score) is somewhere in between the conceptual and the material level. In the case of
concept-based art, we can still think of different conceptual levels, with a number
of idea-based tools (idea generation, idea transformation, refinement, deduction, in-
duction, contradiction, etc.) that the artist can use to develop the final work. There
are two abstraction levels, and an interplay between them.
The actual material level may also change in the course of the process. First I may
work with an interplay between concepts in my head and sketches on paper as the
material form. Later, when I am content with the sketches, I proceed to a level where
the concept in the head, as formalised by sketches, interplays with the final material
medium. Maybe any differences in degree of abstraction between representations
would suffice for a creative process, and the transfers between them account for the
complexity of the process?
8.4 Implications for Computational Creativity
Many experiments in computational creativity have been implemented within the
traditional artificial intelligence (AI) paradigm, using techniques such as symbolic
reasoning, knowledge-based systems, statistical models and heuristic search. They
usually operate within a restricted domain, and the form of the search target is of-
ten strictly defined—a solution to a well-defined problem, a postulate that matches
given data, etc. (for a couple of examples, see Lenat
1983
, Lindsay et al.
1980
).
There is an awareness of these problems, and one proposed solution is to add meta-
level reasoning to affect the process and domain itself (see e.g. Buchanan
2001
).
However, the approach at that level is of the same formal nature as the previous one,
equally distant from how we think, and from the complexity of real life. And the
tasks chosen for modelling are often of a scale that would not be considered partic-
ularly creative if performed by humans, such as the harmonisation of a Bach-style
chorale (e.g. Ebcioglu
1988
; see Papadopoulos and Wiggins
1999
for an overview
of similar projects). They are reasonably complex search processes, yes, but more
like optimisation processes than an exploration to extend our conceptual world. The
form of the solution is known beforehand, and it will give us no surprises.
When going through the AI creativity literature, there is a lack of attention to
process as a source for novelty and complexity. The AI approaches are mostly based
on logical analysis of the concept of creativity and novelty, and not how a human
artist goes about when creating something, at least not the creative processes I can
observe in my own artistic practice. As an artist, I seldom know what I am looking
for. Sudden ideas are often related to the domain I am working in, but I do not know
exactly
what
idea I am searching for. Coincidences play a major role in triggering
specific ideas and in shaping more complex creative output.
