Negative space in patterns of data

One of the concepts with which I have been lackadaisically toying over the past few years is that of negative space, the information that is around and between an object or idea itself.

The classic example is that of vase faces.

Click to enlarge.

This pairing of pictures makes a point that contrast is an important element of negative space. It is far easier to see the faces on the black and white vase on the right than it is to see them on the lower contrast vase on the left. But same shape, same negative space.

The artistic idea of negative space spills over into other fields, particularly data. Looking for the information that isn't there. The classic analysis in this category is the US Air Force analysis in World War II of bomber damage with a view to identifying where bombers needed to be architecturally reinforced.

The story goes that they examined all the returning bombers and mapped where the AA and bullet damage which had occurred, yielding the following illustration.

Click to enlarge.

The natural inference was that the wingtips and central fuselage mass needed reinforcing.

Until Abraham Wald pointed out that the analysis was marred by survivorship bias. They were examining planes which had survived, not crashed. That biased sample had an implication. The weak points were not the wing tips and center body mass. That is where damage was incurred and survived to return. The weak points were those locations where there was no damage - cockpit, first engine area of the main wing, the long fuselage to the tail.

That is where damage occurred which caused the planes to crash, i.e. not return for analysis.

Once pointed out, it makes sense and is obvious but it is hard to see initially.

What is the data that is not there telling you?

You can go down a rabbit hole in trying to see negative patterns, looking for what is not there. Knowledge only has utility and value to the extent that it helps survival. Spending too much time looking for patterns where there aren't any and looking for patterns in negative data is the reverse risk of too heavily relying on rules of thumb and heuristics.

Which is why, despite being intrigued, I am also lackadaisical in my search for negative patterns.

However, one of the easiest ways of cultivating attentiveness is by watching the silhouettes created by shadows. Not only are shadows negative patterns, they are ephemeral ones. It also helps cultivate situational awareness which is useful.

Watching for the play of light gives small pleasures. For example, this display of rainbow colors as sunlight refracts through crystal to scatter color on the carpet.

Click to enlarge.

Nothing significant about the observation. Just a minor aesthetic pleasure.

Here is a negative space example though.

Click to enlarge.

What is that? A Christmas decoration? An angel on a stick? An image of a moth?

It is sunlight, playing through a lace curtain onto a white wood hamper cover.

The lace curtain is a geometrical pattern, not an image of anything. And of course, this is the inverse of the lace. The light areas are where there is no lace in the curtain. The image is essentially a type of Rorschach exercise. We see what we wish to see from a pattern without purpose. Matches Adam Ferguson's description of emergent order:

The result of human action, but not the execution of any human design.

What useful knowledge does this yield? Nothing. In this instance. As far as I can tell.

But as I practice looking for the negative space in data, I do think it has heightened my general awareness of the issue and I know that there have been a handful of instances where it did yield insight. Whether the effort is worth the return, I don't know.

But while searching for negative patterns, it does give you the occasional pleasure of seeing angels where there are none.