Learning through creativity : how creativity can help machine learning achieving deeper understanding

Abstract

In this paper, I address the difficult task of analysing the nature of creativity by suggesting a more objective way of defining it. In particular, I propose a minimal account of creativity as autonomous problem-solving process. This definition is aimed at providing a baseline that researchers working in different fields can agree on and that can then be refined on a case by case basis. Developing our insight on the nature of creativity is increasingly necessary in the light of recent developments in the field of Artificial Intelligence. In the second part of the paper, I discuss how an investigation on the main features of human creativity can support the advancement of machine learning models in their current areas of weakness, such as intuition, originality, innovation, and flexibility. I suggest how methods such as modelling the human brain or simulation can be useful to extract the main mechanisms underlying creative processes and to translate them to machine learning applications. This can eventually aid both the development of machine learning systems that achieve a deeper and more intuitive understanding and our exploration of human creativity.