Ten-Minute Talks: Art and Science

"All models are wrong, but some are useful.”

Attributed to George Box, it's a common sentence in my line of study, oft repeated at conferences and in introductory classes. But I think the spirit of the quote can extend further, linking scientific methods and models to the practice of making art.

But before I drone on about modeling, let’s define what we mean by “art.” This seems to be a passively polarizing question. On one hand, you have plenty of people who argue that the abstract stuff isn’t actually artistic. This comes along with a snide comment like “anybody could do that.”

And on the other hand, there’s not a lot of creativity that in merely copying something down exactly the way it looks, regardless of how technically impressive the skill is.

Fundamentally, art is a model of something else. For things like sculptures and realist paintings, that’s clear. But as art becomes more abstract, so does its link to the world. The artist uses their medium to express only particular aspects of reality. This capturing isn’t just what’s in the painting that but what’s left out. Art requires stripping away elements of reality deemed irrelevant to the focus of the work. Whether it’s by drawing without color, sculpting without definition, or abandoning realistic dimensions, art is an inaccurate representation of reality. In what’s left, the artist has focused on a narrow view of the world. This connection between reality and the narrow view is where the artist develops and executes a creative vision. And it’s the workshop where artistic meaning and beauty are developed.

So how does this relate to science?

Every theory in science makes similar assumptions to concentrate on a particular truth, much in the way an artist only depicts the aspects of reality that matter to them.

Let’s take, for example, queuing networks. Queuing networks are a way to capture the way things move within a system. The canonical example is a grocery store, where people wait in a line (the queue) to check out at a register (which is the “service”). The goal of queuing networks is to model this process with some assumed arrival (people getting in line) and service times (people’s groceries getting checked out), capturing the uncertainty (through probability) that comes with everything in our lives.

The final version of the model is just a set of nodes and mathematical values. These nodes don’t consider what brand the grocery store is or what their floor looks like. In fact, just looking at the final model, you could probably come up with some alternative ideas of where it came from. Maybe it models people at a drive-through. Maybe it’s the line at Chipotle. Maybe it’s pickup basketball.

In the process of creating the mathematical model, the modeler is abstracting away from reality to strip away the things that aren’t relevant to the goal. If it hasn’t clicked yet, this is the same practice the artist does. And the final model, much like a piece of art, is a deep but narrow view of reality, where the modeler has developed meaning through abstraction.

To do this, we have to make abstractions into a mathematical model.

Now there are plenty of arguments that you can make about why art and science are, in fact, different. Where science is often directed to a strict purpose and follows stricter rules, art is more free both in motivation and in practice. But at the end of the day, both employ the same abstraction to create meaning. Maybe, with a little thought, we can also view science’s abstractions as beautiful.