Entropy and Stability in Painting: An Information Approach to the Mechanisms of Artistic Creativity

Abstract

The author presents a method for identifying the sequence of stages of the creative process in painting, specifically an artist's decisions about such parameters as genre, composition, color structure, etc. The method is based on the calculation of the coefficients of informativity relating to the distribution (and hence, entropy) of the artist's works over the above parameters. Using this method, 240 paintings by three famous Russian artists (Pavel Kuznetsov, Constantin Korovin and Ilya Mashkov) were studied. Both the common features and the individual peculiarities of their creativity are revealed, including the "integral stability" of each artist's creative process and its changes over time.

When I established the contrastive lines of my composition, together with the contrast of colors in my palette—my painting was ready.

—Eugène Delacroix

When considering the creative processes of artists (composers, painters, poets, etc.), one of the most fascinating tasks is to ascertain the sequence of the stages of this process. For example, we can presume that some painters attach great importance to the content of their oeuvres. So, first of all, they decide what should be depicted—for example, a landscape, a portrait or a still life. Only afterwards does such a painter decide upon other features (e.g. geometrical sizes, proportions and color structure). I think that this was the creative process of most "classical" painters (e.g. Velázquez).

However, other sequences of decisions are possible: a painter's most important decision may concern, for example, certain color features. Hence, such a painter first decides what set of colors should prevail, then the painting's absolute size, then its proportions and finally the content of the work. I think that this was the creative process of Matisse, for example.

Of course, in no case do I use the terms "to decide" and "decision" as designations of conscious procedures; these terms are used in a rather broad sense. It is well known that the processes of the unconscious are usually very important for creativity. Nevertheless, when creating a work, each artist always chooses quite definite gradations of certain parameters, as if these gradations resulted from decisions.

Naturally, an artist can use one type of strategy in some works and other types when creating other ones. However, we may presume that in most cases each artist possesses a more or less uniform hierarchy of decisions when creating the majority of his or her works. At least, such a uniform hierarchy can be valid for a definite period of creativity. Of course, this is only a supposition, but it is confirmed below by empirical data. So, let us suppose that the above hierarchy does really exist, and we wish to ascertain it.

What methods can be used for this purpose? Sometimes a researcher can rely upon an artist's statements, memoirs, etc. Sometimes the impressions of eyewitnesses (friends, relatives, etc.) can be used. However, it is desirable to derive a more objective method, which could be used as an auxiliary to more traditional studies of creativity.

The most suitable method is to use features of the results of the artist's creative process (i.e. parameters of musical oeuvres, paintings, etc.). The structure of statistical links between these features contains much information about the creative process. The goal of my investigation was to reconstruct the sequence of stages in the artist's process, using statistical data characterizing features of each artist's works.

I describe below a method I have derived from the framework of the so-called information approach [1]. I applied this method to study the creativity of three famous Russian painters: Pavel Kuznetsov (1878-1968), Constantin Korovin (1861-1939) and Ilya Mashkov (1881-1944).

Using Entropy to Measure Statistical Links

In the framework of the information approach, the key concept is entropy. It is a measure of the disorder inherent in the system studied, that is, the distribution of its objects (elements) over certain parameters. The formula for the entropy H_X of the system, consisting of objects distributed over different gradations of a certain parameter X, is well known...