So the way this article uses the concept, a web of trust has the purpose of being able to verify whether a pubkey is associated with a human. Like key signing parties. That of course makes sense and it could be done by a face-to-face meeting, or you could be pretty confident on other grounds that you are dealing with a human. But if that's the idea, it's unclear whether there is any special technical solution required. Maybe the ability to filter events by follow-distances, if you think that those you have verified to be human are themselves reliable verifiers of other humans. It seems at any rate very different from the popularity-score concept.