The idea for this entry came from some very interesting comments on my previous ramblings about the nature of tagging. It seems to me that there are a lot of strange notions out there about the relationship between concepts, words, and tags. Let me first give you my simple view on tagging, which I think is probably right.

1. Our heads are filled with concepts. We don’t really know what these are like, and philosophers and psychologist (and, unfortunately, computer scientists, mathematicians and witch doctors) will argue for a long time about the best way to characterize them. Nevertheless, we know that children have them before they learn to speak. For example, the psychologist Liz Spelke has done fascinating work on the nature of “innate concepts” in infants. I think animals also have a pretty rich conceptual system.

2. Humans (but not animals) have an instinct to communicate with language. Language provides words and grammar to communicate the concepts in our head. I won’t talk about grammar. Words are pretty much arbitrary sounds that we use in speech. Words have a sound that we learn. They also have some syntactic properties that tells us how they should be used in sentences. So I am thinking about cats. I can say “I am thinking about cats”. “Cats” is the sound I use to talk about the cute furry things I am thinking about. If I was talking to a Hungarian about cute furry things I would say “macska”. It pretty much doesn’t matter which sound I use as long as the other person also knows it. But I wouldn’t say “I catted home today” because I know that the word “cat” is a noun and I shouldn’t use it as a verb. I might also suspect this is the case because nouns tend to be about things, and we can’t use them in place of actions.

3. So words are pretty handy. So now I go to del.icio.us and find a web site which I think is about cats. What should I tag it? Geez. Hmmm.

So there you go. Of course there are some messy steps here. There always are. How do I decide the site is about cats? We don’t really know, but this popular blog suggests one mechanism by which humans (and birds) can make quick intuitive category decisions. But once I decide it is about cats, why wouldn’t I tag it “cats”??? Of course I might add other tags. I can add adjectives which are usually descriptors of some sort. “cute” comes to mind.

I don’t believe that adjectives form categories the way nouns do. So I don’t think there is a category of “cute things” or “big things”. Nouns have pretty fixed meanings, but adjectives like “big” are practically meaningless without the nouns they modify. A tiger is a big cat, but is it a big thing? So of course tags which are adjectives are a lot more personal than nouns, and don’t represent categories.

There are of course other issues. There is ambiguity: someone might label a bulldozer a “cat”.  Or, maybe someone thinks the site is not about “cats” but “obsessions”.

But neither of these issues is a problem for my simple view. Ambiguity occurs because we chose not to have enough words. We have the same sound for two different categories. But the process is still the same: “categorize then name”. The second issue is also O.K. as far as I am concerned.

Tagging works because people are free to go with their “gut feeling” on what categories things belong to, and they are free to use the words they have always used to describe those categories. And then they get to put some spice into the mix with some colorful adjectives (emotional terms?), proper names, and verbs (action words?).

Simple! And what is the alternative? I have not seen one. All I have seen is suggestions of magic …

Not one day after my posting that tagging is really at least a little bit like categorizing, I discover this blog
in which there is a link to presentation by Thomas Vander Wal himself, in which presentation Thomas says:

I guess the first part of this disagrees with my claim that tagging does involve categorizing. Although the phrase “not so much” is a little bit confusing. Maybe they are categorizing a little bit even by Vander Wal’s reckoning?

The second part I must admit I don’t understand. What is it to “connect items” and “place hooks”?  Does it literally mean to say that I can “connect” elephant with subatomic particle? What is the mechanism for this? How does it work? Does an arbitrary connection really lead to “tagging that works”?

Tagging has been around a while now, but we still don’t seem to know what it is. Some people argue that it is the complete opposite of categorizing, while some believe that tagging shares much with categorization. I go even beyond that, and I have argued in several publications that rich structural information can be extracted from tags, if we know where to look.

But why does it matter what tagging is? Can’t we go on tagging even if we don’t really know what we are doing? (Works well in the rest of our lives!)

Well, yes, we could go on tagging. The question is what to do with the tags once we have them. If tags are really not categories, and if the extreme view is right that they are simply some sort of completely individualistic associations, then it seems to me that we can’t do ANYTHING with them across the user base. If on the other hand there is some agreement between users about basic categories at least, then we should be able to aggregate them accross users and use them in interesting ways.

The fact that we already do aggregate them and interesting patterns of popular tags can be found, suggests to me that the extreme view against tags as categories must be wrong.