Parsimonious Art

or objectively measuring art

Rather than going for optimizing “aesthetic value”, what about “art” that tries to optimize using as few “resources” as possible (no I’m not talking “minimalism” here, bear with me)?

I’ve studied statistics, where a statistical model with fewer parameters is preferred over a statistical model with more parameters. Try to explain a phenomenon with as few as possible variables.

What happens when this is applied to art?

What makes up a cat?

It possible to define some physical features of a cat. We might distinguish a cat because it has 4 legs, a tail, cat whiskers, cat eyes and cat ears. Perhaps by it’s behavior.What really defines a cat?

As for drawing… how many strokes would you need to convey the concept of a cat? On which aspects of a cat do you focus? It might even be culturally determined what would “work”.

If we consider parsimoinousness in terms of “how many strokes needed”, it is required to figure out what the minimum need is (e.g. in strokes) to define what makes up an object (e.g. cat).

I guess it would definitely take skill to be able to do well in these kind of challenges. The fun part is that it certainly would be a way to measure art objectively (are we even measuring any kind of artistic value here?).

Parsimonious art

The goal would be to define a cost function for art, which should be tried to be minimized.

For example, when wanting to draw a house, the goal would be to use the least strokes possible. How many strokes are required for a majority of people to still recognize it as a house?

Note that even in the definition of an objective, it might just be the goal for someone to guess that 2 strokes would mean a house.

House in 3 strokes?

Minimizing functions for parsimonious art

One definition might be:

minimize(score) = num_strokes * percentage_correct_guesses

Another might be to minimize the total line length in units (there is a slight difference, in drawing more detail, some strokes might be done “longer” to allow smaller strokes in between)

minimize(score) = total_line_length * percentage_correct_guesses

In the house example above (line below and 2 times roof lines), this would be:

(1 * 4) + 2 * sqrt(2^2 + 2^2) = ~9.657

Another minimizing objective might be surface, which could be defined as the surface between (min(x), max(x)), (min(y), max(y)) where x and y refer to the x and y-axis.

minimize(score) = surface * percentage_correct_guesses

In the house example above, this would be 5x4 blocks = 20.

Perhaps color could also play a role? E.g. the number of colors used.

Need for a community

To make a fair assessment, it would make sense to have a community around parsimonious art, where users try to be the best at most accurately guessing what has been drawn. If the drawing is bad, the majority might be wrong, so there should be a “high” score (which is bad). Oppositely, if a lot of people are agreeing on what has been drawn, this score would be low. Thus, it becomes weighing at one hand to minimize number of strokes, while at the other hand trying to maximize user agreement.

To battle users who are guessing wrong on purpose, it is possible to give lower weight to the votes of users who are not agreeing with the majority/correct picture often.

Conclusion and beyond

It seems like it could be an interesting objective to find parsimoniousness art.

Another interesting aspect might be how many people guess the super class rather than subclass. For example, when you would draw a cat’s whiskers, people might guess “cat whiskers”, while others might guess the super class “cat”.

Questions for you:

  • Do you think people would correctly guess “house” on this 3 stroke example?
  • Can you do it in 2?
  • And how many strokes do you think you need to draw a cat?
  • What defines a cat artistically?
  • How many strokes for an emotion?