Human beings create art in everything that we do. Consider food: we could run after wild animals, club them with a stick, and then consume them raw. In ages past, perhaps that’s exactly what happened. We have tuned that basic function, that of eating, into art. It’s hard to deny that Beef Wellington or Crème Caramel are art. Consider housing: it is possible that humans once lived in caves, but even then we decorated them. Now we make elaborate structures decorated inside and out, with amazing colors and metal and glass and a huge number of aspects that are not really needed to house us. The same is true of clothing, transport, every aspect of human life. We are the ultimate jackdaws.

There are words that are very hard to define with any precision, so when we think of art there could be many images in our minds. That’s okay, and it introduces the concept of randomness in art at the highest level: not everyone thinks of art in the same way, and some artworks will have success and others not. It’s not possible to design success into a work.

For something to be random it must be unpredictable [5]. There are degrees of randomness as well. The weather is only partly predictable, for example. The exact time of the arrival of the letter carrier at your door is unpredictable, although it is probably known approximately. So it is with art. There are many aspects of art that are unpredictable to a greater or lesser extent, and that unpredictability is often due to the fundamental unpredictability of humans themselves. Humans are not predictable as individuals, and as art is a fundamental human activity, art is unpredictable too.

It is really not possible for a human to reproduce a drawing or painting precisely. We can get close, but precision eludes us, and is the domain of the machine. But that is only one aspect of randomness in the artistic realm. It seems to be a hierarchy of considerations, beginning with the most complex one: the thought processes of the artist.


A fascinating example of artists having distinct renderings on a single subject was presented by Monet in his Haystacks series. The subject was a rural scene containing some hay stacks, trees, and hills all observed from the same location. Monet seemed to be attempting to show the differing illumination and mood at different times of day, in different seasons. (Figure 0.1). This means that the lighting, season, weather, and most other environmental issues were in variation, and the differences in the paintings would be because of how the artist saw the subject at those times and in those moods. As can be seen in the figure, they are quite different. Distinct parts of the image have more or less detail, and the overall effect is not simply a matter of illumination differences. Why? Because each artist always has a world view that dictates how they will transfer a visual scene into a rendering. That world view is a high-level component of how art is created, and definitely has a lot of apparently random components. In fact, it’s possible that if we knew about every aspect of an artist’s life and understood how that would affect their personality that it would be predictable, but there are simply too many factors to consider. It is, to us the viewers, unpredictable.

Interestingly, when one asks a collection of art students to paint a tree, they generally produce paintings that are startlingly similar to each other, depending on their skill. Inexperienced artists tend to render accurately as they understand it. Artists appear to allow their world view to affect their work as they gain experience, and then they are said to have developed their own style.

The Artist’s Physical Location

The location from which the artist views a subject is often referred to as the point of view . This refers to a physical point of view, not an intellectual one. There are many places from which an artist may view their subject, and that choice dictates much about the visual presentation of the scene. The scene, as has been noted, is three dimensional, and the artwork is flat. The subject has, from the point of view of the artist, a front and a back, top and bottom, and side. These are all relative positions.

The location of the artist in the scene is unpredictable, and so is a random parameter of the work. The geometry of perspective defines most visual aspects of a rendering, and this mainly depends on the artists physical location and orientation. A small change can make quite a difference. In Figure 0.2 the change is one of orientation, but position matters just as much. As objects get closer they appear larger, for example. The location and orientation of the artist, the viewpoint from which the rendering is accomplished, is a random component of the work. Of course the cubists changed a lot of that, but not for everyone [1]. It may be instructive to see what they did, though. For instance, objects are sometimes painted one on top of the other instead of being in adjacent positions in the third dimension. Multiple views of an object can be unfolded and flattened so they can be seen simultaneously. Distance could be rendered as colors rather than with size changes: reds and browns for foreground objects, and blues for the background.

Consider Picasso's Girl with a Mandolin. Here he did something new: he represented a figure as a series of small transparent planes or polygons. All perspectives were visible at the same time. Each polygon has a distinct orientation in space and so has a corresponding distinct shading, given the lighting.
This shading does not correspond with the surface of the face being represented, but has a random aspect that yields a random perspective. An interesting question is whether Picasso, when presented with a black canvas and the same subject, would have created the same portrait involving the same polygons in the same orientations. Likely not. The result would have a similar overall appearance but would be, in detail, unpredictable.

Explicit Randomness

A Dadaist, Jean (Hans) Arp was one of the first artists to make randomness and chance part of his work. He generated the form first, and tried to reduce the intervention of the conscious mind. For example, he dropped bits of paper onto a surface to see what would be produced.

The Dadaists were excited by randomness as a partner in creation. Although it was a protest movement and rather short lived, the influence did carry over to successive schools and styles.
Artist Paul Klee, only a casual follower of Dada, used randomness too, but in a more intellectual and structured way. as well, but in more `abstract" terms. A component such as a line, he thought, had “no goal”. The artist could not avoid thinking of the line in a certain way, as having a purpose and destination, but the line itself did not. Thus, the artist has to give up some control. It is ‘an active line on a walk, moving freely without a goal. A walk for walk’s sake’. This sounds very much like a mathematical concept called a drunkard’s walk or random walk in which a line starts at some point and moves in a random direction at each step.
Some of Jackson Pollock’s work has been criticized at times for being random. His famous drip style has a visually random appearance, although Pollock claimed close control over the creation process. He would drip paint onto a canvas from a height, controlling color, length, and speed of the stroke [6]. It might be said that his work was semi-random, sometimes careful but sometimes subject to the whims of gravity and viscosity. Perhaps a good description of his would be ‘spontaneous’. (Bluepoles Source: )
Later still I the 20th century, Fred L. Whipple devised a method that he referred to as stochastic painting. The word stochastic means “involving chance, or a random variable.” Stochastic paintings have colors and shapes that are organized with random components, but in an explicit way. They are designed to be random. He may be the first truly generative artist, because his technique requires that integration of design and chance, and he insisted that a stochastic painting be based on a set of rules, which we would later call an algorithm. Stochastic painting is interesting partly because the outcome of a specific algorithm is not always predictable. Until the work is finished, it may not be possible to completely visualize it. And, of course, two painting created using the same rules can be quite different. One could be terrible, and the next could be amazing.

Which kind of rules would Whipple use in creating a painting? Here’s an example [8]:

1. The first pair give and on a canvas coordinate system for the starting point.
2. The first of the second pair, taken as a decimal of , gives a direction from the starting point; the second, multiplied by a unit distance, say a centimeter or half an inch, measures a distance in this direction.
3. From the end of the first line the first number of the next pair measures a distance; the second, multiplied by , measures an angle turned counterclockwise from the tip of the previous line.
4. Successive lines are developed by successive number pairs from the ends of the previous lines or from the outer sides of closed areas.
5. We now must have a rule for closing the areas. I first tried a rule that produces areas that are all triangles or polygons with no internal angles greater than . I chose to join the figure at the end of a line when any projection of a line was pointed towards the originating side of the polygon. This rule leads frequently to several lines radiating from a point, which gives some sense of three-dimensionality to the final painting. (Fig 0.8)
6. At the edges of the canvas I first adapted the simple rule of extending the line by equal-angle reflection.
7. When the canvas is completely covered, the choice of colors can be made by successively numbering each closed area by a number taken in sequence from a random-number table. The nature of the painting can be quite affected by ruling that contiguous areas may or may not receive the same color. In Fig. 0.8 I chose to eliminate contiguous areas of the same color thereby ending up with colored areas all of polygonal character.
8. If the tubes of paint are numbered successively, in any order, ten random numbers distribute the ten colors among the numbers from 0-9.
9. The remainder of the operation, as in any number painting, permits the painter to choose textures and shades at will. Or, if he wishes, he can mix a certain amount of white with the paint for each area by means of a second random number in each area.
This is, by anyone’s standards, an algorithm. It underlies modern generative art.

Random Effects Create Backgrounds and Imply Shape

Scenes have objects that are of concern, thought of as foreground, and these are placed in a context that is less important, thought of as the background. We can argue about whether the background is more or less important, the merits of negative space, and other concerns, but the fact is that the background does not always need the same level of detail [4].
Consider the painting in Figure 0.9. The background is simply empty. It could be left as a solid colour or it could be shaded. Here a pencil; shading is used in contrast with a solid white, and whichever one seems more interesting, the shading of the background is random, at least in detail. This is a common effect in pencil and pen/ink renderings. Pencil shading is always random, since the way graphite is deposited onto paper has far too many variables to control. Some control is possible, and so the harder we press on the pencil the darker will be the shading. The details of the pencil texture are random.
Pencil textures can also be drawn as many small objects, lines, or curves. As these small objects are closer to each other they appear darker, and so they are used for shading to imply shape. If the objects are drawn in a regular fashion the effect is lost.

The cat and the sphere in Figure 0.10 are shaded using “+”, “-“, and “/” characters where their density increases with darkness but actual positions are random. This is typical of the shading method. This is taught by art teachers as a standard method, and there are many random features that can be used for shading: cross hatching, line density, dots, stippling, cross contours, scumbling, smudging, and more.

Randomness Can Take the Place of Reality

As we’ll see in Chapter 3, the real world is an incredibly complex place full of events that are related to each other, but we rarely know what those relationships are. As a result, a lot of the world seems random. This is why humans invented statistics – so we can have at least some idea of how complicated processes may turn out. When will the next bus get here? I don’t know, but we would expect it to be here in the next five minutes. More or less. Approximately Just about. These terms reflect our lack of detailed understanding of the Universe.

This affects art in many ways. When rendering a scene for reality we are usually content to render what we see, and the randomness of the world lends the result a sense that it is real. When rendering an imaginary scene, we have to introduce that sense. Trees don’t grow in neat rows unless they are planted. Falling leaves create random seeming overlapping patterns. The shadows made by curtains flapping in the wind don’t make a regular motion.

The tree on the left in Figure 0.11 is a watercolor based on the artist’s imagination, not a real tree. So is the one on the left. Which one appears to be more real? While one may prefer the tree on the right, it is clear that the other is more typical of a real tree. That’s because real trees have a random-seeming collection of features. The location of branches is unpredictable, as is the number of branches, their size, and so on. To make something seem real we need to introduce chance elements into the visual composition. No two snowflakes are the same, they say, and that applies to a great many things. Like trees. When a human sees too much simplicity, they see something artificial.

Complete Randomness Is Not Interesting


[1] Kristin Brenneman (1994). Chance in Art.
[2] Michael Challinor (1971) Change, chance and structure: randomness and formalism in art. Leonardo 4 (1971), pp. 1-11.
[3] Jiayue Gu (2017). Randomness and Control in Contemporary Code Art, Honours Thesis, The Research School of Computer Science,Australian National University
[4] Carl Lostritto (2015) The Value of Randomness in Art and Design, Fast Company Oct 19.
[5] Stanford Encyclopedia of Philosophy (2018) Chance versus Randomness,
[6] Kenny Verbeeck (2006) Randomness as a generative principle in art and architecture. Thesis (S.M.)-Massachusetts Institute of Technology, Dept. of Architecture.
[7] Walt Disney Productions (1958). Four Artists Paint One Tree.,
[8] Fred L. Whipple (1968) Stochastic Painting. Leonardo, vol. 1 no. 1, 1968, pp. 81-83. Project MUSE,