For many years the meetings of the scientific community working on Computational Creativity started and ended with long hours of trying to provide a workable definition of creativity. As these discussions tended to be of a circular nature, and they would rarely result in any relevant insights, the community learnt to steer away from them, and focus on more practical issues related with the simulation of the particular behaviours that we wanted to reproduce. This has lead over the years to the creation of a substantial work of empirical work both on the construction of computational models that simulate creativity and of qualitative and quantitative methods for evaluating creativity of such systems. In that sense, there is a positive aspect to the decision of staying away from definitions of creativity.
There is, after all a precedent in the field of Artificial Intelligence, tha faced in its time similar problems and adopted similar solutions.
However, there seem to be recent attempts to reconsider that decision in certain fields of AI. A recent post by Menno Mafait to the LinkedIn forum on Natural Language Processing addressed this point in the following terms:
"AI scientists decided 60 years ago to develop techniques on AI and knowledge technology (including AI programming languages) without defining intelligence (and meaning) in a natural and deterministic way. In this way, AI and knowledge technology have no foundation in nature. And the text books used during my education agree: AI is a simulation of behavior.
AI is based on clever engineering. And knowledge technology is based on applying smart algorithms to keywords, by which the natural meaning of non-keywords is ignored. In this way, humans are still the only source of intelligence and meaning in AI and knowledge technology.
No fundamental results can be expected – and no fundamental problems can be solved – as long as the laws of nature involved with intelligence are ignored: Intelligence and language are natural phenomena. Natural phenomena obey laws of nature. And laws of nature are investigated using (basic or) fundamental science, not be using (behavioral or) cognitive science. So, in order to have fundamental results – and solving fundamental problems – we have to redo most work done over the past 60 years, using fundamental science by defining intelligence (and meaning) in a natural and deterministic way first."
You can read the original post here:
From my point of view, the benefits of the empirical work far outweigh the possible long term advantages of going deep down to define the subject matter in a natural and deterministic way. But maybe there is room for considering this alternative path to knowledge.