RE: [huml-comment] Request for a motion on PC-33 -Section 4.4.6-r ace

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

That's precisely the point.  Either the individual 
has to assert their membership in a named set, or 
some set of observable even testable properties has 
to be used (intension or extension).   

That is exactly why I am providing 
this example.  It is real and used in public safety 
databases in America.   Most of us here realize that 
such codelist value assignments are unreliable 
without more information on how the selection is made. 
As you show, the genetic information is almost worthless, 
so otherwise, it is just a sign for which properties 
have to be declared to enable selection.

Real example.  A friend of mine is from Honduras.  Her 
mother is Honduran; her father is Brazilian.   Her 
ethnicity assignment is Hispanic.  On her birth certificate 
and her driver's licence, her race is "white" and that 
was a big surprise to the family of the North American 
white family she married into.   Unbeknownst to them, 
her Brazilian biological father is black, although 
the father who raised her, her mother's second husband, 
is white but Hispanic.

Messy.   Why would they collect such obviously 
messy data?  Typically, for statistics applied not 
to individuals but to large groups.  To clean that up, 
one could create individual profiles with more information 
useful in making the selection, but I believe the 
expense of that will also point out the relative 
uselessness of the race classification because based 
on observable characteristics, it is fuzzy and leads 
to misclassification (and thus miscommunication) in 
individual instances, and otherwise, requires the 
individual to select the value.

IOW, as signs go, it is fuzzy and to defuzz it, 
it is expensive.  Something "The Fuzz" should 
become aware of. ;-)

len


From: Emmanuil Batsis (Manos) [mailto:mbatsis@humanmarkup.org]

Bullard, Claude L (Len) wrote:

> Race:
>  
> A = Asian
> B = Black
> I = Indian
> W = White
> U = Unknown


What if my father was half Asian and half White, and my mother half 
Black and half White; what would I be?
Given that this attribute is just of type xs:string and not an 
enumeration (not that such a constraint would help) does not really 
help. The information contained is to be unpredictable at best.

Putting ethical arguments aside, this attribute does not serve much. 
Now, if a complex type could trace races up my bloodline, that would be 
something. Yup, that would be pretty objective...