OSPF uses a thing called routerID. When not explicitly specified, it uses the IP address of the loopback interface. Without it, it falls back to using highest IP address of any adapter in up state.
Latest method can create problems. If an adapter goes down, OSPF has to change its RouterID, taking down all learned routes.
A loopback never goes down and thus is better choice. (Manual configuration of router ID is alternative)
The loopback also gives an extra pingable address, so you can still test connectivity (ping) to router regardless of which of its interfaces is up.
I agree most KBs are just cookbook recipes. Put together these ingredients and you endup with a nice soup. But without a real explanation of its inner workings.