Jeff Atwood, the founder of Stack overflow.com and of Discourse.com, two of my favorite sites, tweeted recently: "I'd even go so far as to say folksonomy/tagging is the #1 overrated concept in the social media world." This followed up a tweet expressing his deep ambivalence "about tags as a panacea based on my experience with them at StackOverflow."
At this point, I don't know anyone who suggests that tags are a panacea. But are they useful?
Yes. Often. And often they're useful alongside a taxonomy of some sort; there's no contradiction between providing a standard set of terms by which users can navigate and enabling them to also use their own.
In fact, in the bashing of tags back and forth, we should remember that even if they are not a panacea, they upset a very old applecart. It was taken for granted for a couple of millennia that the universe is neatly ordered into a hierarchy of ideas, and for many centuries that those ideas are nested in a tree-like shape. Further, it was believed that there is a right and true way to order ideas, and that it was up to experts to figure that out. This reflected the limitations of the physical world in which things can only be organized in one way at a time: If you organize your books alphabetically by author, you can't simultaneously organize them by the date you purchased them (unless you have surprisingly obsessive buying habits).
In this world in which authorities controlled how things were organized, tags permitted individuals to organize stuff—theirs and others—the ways that made sense to them. The online bookstore may say that Chariots of the Gods is a science book, but you can tag it as "hogwash." Put together all the tags supplied by all the people and you get a cloud of terms. With some processing, this cloud can be turned into a "folksonomy," which will vary based on what rules you apply. For example, you might decide to combine tags with the same stems ("science" and "scientific"), or even to combine tags that a thesaurus says mean the same ("hogwash" and "bunkum"). Or you might not. Classification is a tool, so it all depends on what you're trying to do. Nevertheless, tags were important when they first made a splash on the Web in the early 2000s, in part because they represented a break with some old assumptions about authority and meaning.
Whether or not tags are being used as widely as some of us (including me) thought, the point that they made—we are no longer powerless in the face of how others classify things for us—remains true and valid ... so much so that it has been absorbed and taken for granted.
But aside from the theoretical point tags make, do they work?
In one of his tweets, Jeff points to a set of tags concerning the video game Battlefield 4 that embody some of the main problems with tags: They cover the same topic in multiple ways ("part 1", "#1", "episode 1", "ep. 1") and can be meaningless to others ("funtage"). The same tag can also mean different things to different people, leading to unexpected results when someone clicks on the tag "game" expecting to see items pertaining to video games and instead sees information about shooting pheasants and deer.
But no one ever said that tags were without problems. Messiness comes with their bottom-up, scaled nature. That doesn't matter if you're just browsing photos. It gets in the way when your work requires you to find everything about some topic. That's one important reason that professionally curated environments that support domain-specific researchers don't rely only on tags. For example, a law library would be crazy to rely only on user-generated tags.
Part of the difficulty in talking about tags, though, is that they only exist when implemented in a particular system. For example, Jeff tweeted "we made them work OK over time, had I to do it again, max 3 tags not 5." Every tagging system has its own set of rules and functions. Some systems automatically suggest previously created tags as you're typing in a new one to try to bring focus on a consistent set. Other systems mediate or curate tags. Wikipedia has a rule for its tag-based category system that you cannot tag a page with a concept under which it is clustered; for example, you can tag Of Mice and Men as "Depression Fiction" but not as "Novel" or "Book." Many systems, many rules. If your tagging system isn't working, the problem probably isn't with tags in general but with the way you have implemented them.
Tags work, but imperfectly, which is how anything that works works.