I made use of the fact earlier that humans can learn new coding algorithms. In a way, our innate coding algorithm is universal enough to allow for implementing new coding algorithms. One important subgoal of low-complexity art is to devise good coding algorithms. What does this mean? A new coding scheme may be considered ``good" by a given observer if (1) it does not require many bits to be specified (given the observer's previous knowledge), and (2) many different drawings (satisfying typical specifications) can be encoded efficiently by it. In that case the drawings share a non-trivial amount of algorithmic information, and the coding scheme represents something like a common style. A given coding scheme may be representative for a given artist, which will make that artist stylistically recognizable.

More formally, the quality of
an artistic style or coding scheme *C* may be evaluated as follows.
An optimal style *C*
maximizes
P(C | S),
the conditional probability of
the style, given a set of drawings *S* (defined by
a set of specifications).
Equivalently,
an optimal style *C* minimizes
-logP(C | S) = -logP(S | C) + logP(S) - logP(C).
Since *S* is given, *P*(*S*) may be viewed as a normalizing constant,
and it may be ignored. The term,
-logP( S | C)
can be interpreted as the
information required to compute all elements in *S* from *C*.
The term, *P*(*C*), is given by some a priori distribution on the
coding schemes and depends on the observer.
-*logP*(*C*) can be interpreted as the
information necessary to specify *C*, given the observer's
knowledge.
Thus, given *S*, *C* is optimal (most likely)
if the sum of two terms is minimized:
(1) The information required to compute *S* from *C*, and
(2) the information required to compute *C* from the observer's
previous coding scheme.

The circle scheme is easy to teach, which may be another way of saying that not much information is required to compute it from typical human knowledge. In this case, the second term appears negligible. Therefore, given the drawings presented in this paper, the circle scheme appears to correspond to a ``good" (although probably non-optimal) artistic style.

Back to Theory of Beauty page

Back to Algorithmic Information page