In her Cognitive analysis of tagging, Rashmi Sinha hits the core problem when she points to our fear that we would make a wrong categorization decision.
“We need to consider the overall categorical scheme.”
If the overall scheme is a restricting, inflexible one that does not allow for cross references or “see also” links, this fear is certainly justified. But many hierarchical systems do offer xref links.
Even in my windows I can create a shortcut link from a folder to another folder if I am unsure about where I might look next time. So I can be confident that a suboptimal categorization won’t do any harm.
Many people also seem to be intimidated by the huge “overall schema” that tries to catalog the whole world, starting from monstrous supercategories like “The Universe. Mankind.” and the like. But especially the top-level categories are not at all important if you have your shortcuts mapped from a more convenient view onto the given schema.
Often, the top categories can easily be mapped onto facets of a faceted classification schema (see this blog discussion of a class at SIMS, Berkeley): The top levels of your schema may start with topic, then proceed to chronological coverage, and end up with geographical drill-down (as Dewey DDC does). Or first chronological, then geographical, then topic. Or any other permutation: it does not matter – provided that I have mappings in place. And the same applies to the root categories of the topics. If some traditional schema wants to start with divinity, law, medicine, and then summarize everything else under philosophy, why not? That’s not really a problem.