I recently posted about the Relationship vocabulary, and subsequently (coincidentally too) Clay Shirky wrote a critique, that prompted many responses (see the comments and trackbacks that follow the critique), that prompted his further clarification and critique. In response to Shirky's first critique, I wrote a long letter that I decided to trash and not send or publish, but after seeing and commenting on danah boyd's RELATIONSHIP: Context, Culture, Power post, I felt like I should retrieve some of my letter, put it a little differently, and publish it here.

The background for my comments is that I have worked pretty extensively with controlled vocabularies, I have had to develop more than one extensive classification scheme for people and groups, and I have worked with large databases with information about people and groups (e.g., CRM type data).

From this perspective, I feel like some of the things Shirky and others are concerned about happening (e.g., in the affects of labeling relationships, in the affects of using controlled vocabularies, in the desire to support logical processing of people information) have, in some ways, already happened and been in common use in a gazillion databases—and have been at least somewhat commonly in use in this context for decades.

So, from my experience, I know there are many valid concerns with regards to what happens when "private" (database-specific) contexts become public, what happens when two contexts are merged, and what happens when one context is pushed as being universal.

But, from this experience, I also think it is somewhat "illiterate" to assume that there is an inherent problem with classifying relationships in machine readable formats (and that this problem is somehow universally insurmountable when compared with classifying what Shirky refers to as "external conditions").

In other words, I think the valid issue is about the context in which classification occurs: does it make sense to classify (anything!) in a particular context—and then, specifically, what are the potential affects of classifying something a particular way in a particular context.

This is the comment I left on danah's post:

I like a lot of Clay Shirky's critique and very much agree with your points, but I disagree with Clay's starting point—the assumption that a machine readable vocabulary for classifying human relationships is altogether some new frontier (destined for failure).

There are millions of databases—some running for many decades, that each:

• use a formal and explicit ontology for human relations
• make human relations explicit without changing the nature of the relationship
• use bounded terms for classifying relationships

Vocabularies are themselves situated within a context. Machine vocabularies as well.

I would argue that it is the removing of context for the vocabulary (or, really, the suggesting that a vocabulary could be unbounded by any context, or it could be universal) that is the "wrong path" that leads to the failures you and Clay Shirky indicate.

The Internet / web, in the way it is universal, isn't bounded enough to function as a context in which a vocabulary is effective (e.g., consistent, comprehensive, etc.). So I sympathize with reactions to "universalizing" vocabularies that might be very useful in a particular context, but might seem wrong or weird outside of that context.

To sum up the rest of what was in my letter: a lot of the stuff people are trying to do as a semantic web is already happening within more contained systems that, unlike the web, can enforce things like the use of vocabularies and the semantics of linking. A lot of the classification in these systems already is of people and their relationships. At least some of that classification of relationships works and makes sense in the context of those systems and to the people who use them.

(p.s., hopefully it will be deleted, but I accidentally posted my comment twice to danah's blog. doh! Sorry to danah and everyone who reads her post!)