We have various tools in place to moderate topics, such as merging them. For example, the other day I noticed the topics fitness and getting fit seemed to be used the same way and each had about 130ish followers, so I merged them with the surviving topic being fitness. The software does the right things after merging.
But is there a mechanism in place to inform the people who got merged into another tag that it happened? Because getting that feedback loop established is one of the things that will keep people using "more useful" tags. Without it, they have no reason to change their behavior.
Though it brings up another question of a more emergent nature: did the differentiation of "getting fit" and "fitness" serve a different set of expectations? It would be interesting to look at some sort of lexical decomposition of the post content tagged with each of them to see what words and phrases were often found in one rather than the other. After all, they both must have seemed important at some level to the people that tagged them so.
The easily kind of mad questions I ask. I chalk it up to insanity.
I suspect one of the most important things will be when we can devote someone very knowledgeable to the job. I suspect this is what Medium is doing because it looks to me like they handle topics really well.
Medium has a very strange way of managing their topics. For technical issues they seem to be able to do some fine differentiation. For social issues – well, they seem to suffer from the fact that the development team is politically monocultural. It seems to come across in a very broad aggregation of topics on classically conservative positions and a lot more fine-grained differentiation for progressive positions. Part of that has got to be because their reader base has a very particular political slant and usage has a very particular political slant as a result, but I suspect you are absolutely correct that they have at least one and probably several people internally curating tag clustering and the taxonomy.
I'm not sure it's the best solution, necessarily, because of the potential for human judgment to fall in certain patterns, but it certainly looks to be what they're doing.