I've brought up the McNamara fallacy before in the context of metrics-driven management, something all too common in software engineering. The basic idea is that just because you don't have the ability to easily measure something doesn't make it not important. Something can be vitally important to your success, but difficult to measure for whatever reason.

Let me apply that to what AI is doing to OSS and to common culture.