From: Lee Corbin (lcorbin@tsoft.com)
Date: Thu Aug 09 2001 - 01:05:24 MDT
In the 1970s there were some serious efforts made to create tests that
would have no correlations with ethnicity. All such attempts failed,
I believe. ---Lee
Spike wrote
> Brian Phillips wrote:
>
>> It's a statistical fact that the mean for persons who identify as
>> African-American on most (arguably virtually all) standardized
>> tests are aproximately one standard deviation below those
>> the mean of those who identify as white...
>
> So here is the challenge: suppose we wish to design a college
> admissions test that would *not* favor one subset of the above
> any other. The design constraints are as follows:
>
> 1. The test must be performed in a single day.
>
> 2. The test must not require a test administrator to assist
> the test takers in any way.
>
> 3. The test scoring must be completely objective, i.e, no
> judges or subjective anything.
>
> 4. The test must scorable by machine.
>
> Under these design constraints, what would the test
> look like? Would it not look a lot like the SAT?
> If the blame for uneven scoring is put on differing
> reading skills, is there in theory any way to design
> a legitimate test that does away with reading? These
> are important questions, for Taxifornia is struggling to
> do away with the SAT, claiming that it unfairly favors
> a subgroup. They have moved to replace it with a test
> that includes a foreign language. This improves the
> scores of those who grow up hearing Spanish in the
> home for instance, but further reduces the scores of
> the poor scorers which have no foreign language skills.
> How can colleges legitimately measure and select the
> smartest? Can they? spike
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:09:37 MST