Bias makes ambigious metadata more useful, not less.
A comment on Christina's blog made me think: "And for metadata to be useful, it has to be honest." After reading lots of Lakoff, that just didn't sound right. Thinking about it some more, I think bias is:
1. unavoidable with ambigious metadata (unambigious metadata are things like publication date, author. Ambigious metadata are things like topics or ratings, and can be a lot more useful.) There is always bias. Unbiased categorization is not possible.
2. a value-add. I find it more valuable if a friend that I know (whose biases I am familiar with) recommends me something than if a stranger does the same. Bias ads value to metadata. Or maybe that should be: Known bias increases the value of ambigious metadata.
That's not really what I wanted to say though. More known bias means more valuable ambigious metadata. That's more like it, although I feel I'm still missing something. It's not just that knowing bias is good in a we-can't-avoid-it-so-lets-know-it kind of way, it's that bias in itself is good - it ads personality and flavour to metadata.
Is this something librarians have known all along?