The Shape of the Machine

To change a system, you first have to see its shape. Two properties reveal most of it.

The first is legibility — how readable the system is, and to whom. Systems that can see you in detail while remaining opaque themselves hold an advantage that has nothing to do with being right. Ask which parties can read each other, and on what terms.

The second is feedback — how the system learns it was wrong. A structure with fast, honest feedback can be mediocre and still improve; a structure with slow or corrupted feedback can be brilliant and still decay. When you look at a machine, find the loops: what it senses, how quickly, and who is allowed to act on what it senses.

Everything else in this area — incentives, metrics, and their failures — sits on top of these two.