[Sigia-l] Re: first principals (bias in cataloging & classification)

[Karl Fast]

>> 7. There is no such thing as an unbiased IA. But we have to try
> 
> No - often (if not always) the value lies in the bias. 

Let me focus this a bit:

  Can we be unbiased in classification and resource description?

  Secondly, if we can should we *always* strive for this or are
  there times when we shouldn't?

I'm sure there are papers about this. All unread by me.

Some thoughts.
  
* No bias is often possible:

  A horse is a horse is a horse, of course. And a horse is not a
  cow. What I mean is that sometimes the way we describe something
  is clear and unambiguous (but not always).

* No bias is not always desireable:

  An author search for "Dickens, Charles" in my local OPAC reveals
  TWO authors by that name (The numbers are the birth death/date for
  each, not that it helps me much here):

     Dickens, Charles (1812 1870)
     Dickens, Charles (1837 1896)

  In trying to be unbiased the OPAC is spectacularly unhelpful. Odds
  are I'm interested in the Dickens who wrote "A Tale of Two
  Cities," not the one who wrote "Dickens's Dictionary Of Paris."

* Bias gets stuck in our infrastructure

  Ever take a look at the Dewey Decimal system? It is horribly
  anglo-centric and christian-centric. The problem here is that bias
  is embedded in our infrastructure and once it gets rooted in there
  it's almost impossible to get out.
 

* Bias is helpful in certain contexts

  Bias is certainly helpful from a business perspective. If I'm
  selling books them I want to make sure that popular titles and
  items with high profit margins are promoted better than others.

  The utility of bias in other contexts is contentious.  


Note also that this is about metadata that describes or classifiies
a resource and I am talking about *manually* assigned metadata.

But this may be automatically assigned through some sort of
algorithm. Here things get dicier. Any discussions of bias should
consider the distinction between manual and algorithmically
determined metadata.

These are random thoughts; not properly connected. No time. Gotta run.

-karl