Lowering the costs of metadata

Recently I've been thinking a lot about the role metadata plays in discovering geospatial data and making judgements about fitness of use. It's becoming increasingly clear that the costs of generating standards-compliant metadata (e.g. FGDC, ISO 19115) are comparatively high. Organizations that have no clear mandate to create metadata, create either minimal metadata or worse, none. Even those who are mandated to create metadata (like federal and state governments) sometimes fail to provide metadata. I suspect that some managers of GIS shops go through some cost-benefit calculation that goes something like this:

I produce metadata at my expense.
You, external user, use that metadata to access my data.
Therefore, I incur all costs, you get all the benefit.

So, how do we lower the costs of creating meaningful metadata? I recently picked up the most recent version of Information Architecture by Morville & Rosenfeld. Part of my motivation was to gain a better understanding of the bigger issue of how to structure web-based content to make it finadble and accessible. But more importantly, I want to understand how practices of tagging and collaborative classification schemes assist in finding geospatial data and linking it to other content, which may or may not be geospatial in nature.

Certainly tagging has problems and limitations that have been pointed out elsewhere. One big disadvantage is the lack of support for formal metadata classifications outside of the 'keyword' category. Another problem is how to implement? Data producers will tag their data with certain keywords, but how do we allow the broader community to tag data in ways that are meaningful to them? The Geodata Commons whitepaper acknowledges this difficulty. By temprament, I favor a solution that is open and collaborative and thus, takes advantage of network effects and improves with use. But practically speaking, I'm not so sure this would preserve the value inherent in a controlled vocabulary.