Information Technology Reference
In-Depth Information
When designing computational models of creativity, it is an important question
how to evaluate the success of the implementation and hence how to draw any con-
clusions from the experiments. As I see it, there are two different cases. Either you
make an implementation that tries to generate art that is credible and interesting to
human observers, or you make a minimal model and evaluate it if it exhibits the right
kind of emergent behaviour in relation to its own context, i.e. its limited amount of
data about the outside world.
One could argue that a computational model that produces results interesting to
human observers, having been exposed to a very small amount of human art, could
not be a faithful model. A human with such limited experiences and limited contact
with the outside world would certainly not produce very interesting art. So, the
creator of such a system simply must have put in a lot of his own experience into
the model, consciously or not, in the form of informed design choices.
On the other hand it is difficult to evaluate a minimal model, since it will not be
possible to judge the output as art in itself. The novelty, complexity and interest of
it must be valued in some way in relation to the context and scope of the program.
Creative behaviour may be there, but we may fail to recognise it, since it does not
appear as anything we are used to. Only if a computational model can perceive
and internalise substantial amounts of humanly produced artistic material, combined
with human feedback on its own output over an extended time, can we judge the
output by human standards.
8.5 Final Remarks
I have presented a model of creative artistic processes that is founded on artistic
practice. There is a long way to go before it can be implemented fully in software,
but it is my firm belief that it could help us create more believable artistic results
and behaviour from computational creative systems, and it may form a foundation
for discussion, analysis and increased understanding of human creative processes.
In my own artistic practice as composer and improviser, I can clearly see how it fits
a wide variety of creative activities, and I present it here to be tested by others in
relation to their experiences from art practice and in future computational imple-
mentations.
Though preliminary, the model already provides a framework for analysis, dis-
cussion and possibly emulation of a number of important concepts and phenomena
directly related to creativity:
•
of the relationship between the theoretically, practically and conceptually possi-
ble; between material, tools and ideas;
•
of the relationship between the artist and his tools;
•
of ideas, concepts and generative processes as guiding mechanisms for realisation
of a work;
•
of choices, and how we navigate the space of the possible;
