Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 11/03/84 (WLS Mods); site fisher.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!princeton!astrovax!fisher!david
From: david@fisher.UUCP (David Rubin)
Newsgroups: net.politics,net.women
Subject: Re: Alternative Action
Message-ID: <666@fisher.UUCP>
Date: Tue, 11-Jun-85 09:59:15 EDT
Article-I.D.: fisher.666
Posted: Tue Jun 11 09:59:15 1985
Date-Received: Wed, 12-Jun-85 07:02:51 EDT
References: <1447@watdcsu.UUCP> <657@fisher.UUCP> <1456@watdcsu.UUCP> <663@fisher.UUCP> <1462@watdcsu.UUCP>
Distribution: net
Organization: Princeton University Department of Statistics
Lines: 78
Xref: watmath net.politics:9363 net.women:5722

[">>>",">" = David Canzi, ">>","" = David Rubin]

>>>       ... a second interviewer, spot-checking the first interviewer's
>>> work by reinterviewing a random sample of interviewees.

>>> This scheme fails only if both of the interviewers are dishonest.  The
>>> second interviewer can be (a) borrowed from another branch of the
>>> company, or (b) rented or borrowed from another company, or (c) from
>>> the government.  This reduces the likelihood of the interviewers
>>> getting chummy and covering for each other.

>>First, if it employment decisions could be adequately made through the
>>"spec-sheets", there would be no need for interviews at all.

> Yes, it would seem so.  If the set of relevant criteria was that
> well-defined, it would only be necessary to send job applicants a
> questionnaire...  Your objection is close to a valid objection.  Close
> but no cigar.  See my next posting.
 
I think you have a narrow view of what it is an interviewer does.  See
MY next posting.

>>                                                              The
>>interviewer need not lie on a spec sheet to effectively reject an
>>applicant.

> The interviewer's only way to influence the hiring decision is by
> what he writes on the spec sheet.  How can he do that other than
> by lying?
 
The point of an having an interviewer is to evaluate information which
is not apparent on any spec sheet; such information is often
subjective and therefore unverifiable.  Thus, an interviewer's 
evaluations cannot be proven to be "lies" (willful misrepresentation
of fact), but rather can be attributed to a legitimate "difference of
opinion", even if the interviewer is actually prejudiced against the
applicant.  

>>Second, a single biased interviewer can still effectively reject
>>marginal applicants.  If the first interviewer, e.g., expresses his
>>bias only on those who are on the margin, he will not be evidently
>>abusing the process, and the company is likely to select those
>>marginal applicants recommended by BOTH interviewers.

>>Third, it is not at all unlikely that BOTH interviewers will share the
>>same prejudice, thus subverting the applications of some types of
>>applicants who are well above the margin.

> The second interviewer, if "borrowed" and employed in the way I
> have suggested:
> 1) only checks on the work of the first interviewer, and has no
>    influence on the hiring process
> 2) doesn't know the first interviewer, and so wouldn't care if the
>    first interviewer is fired for falsifying the data
> 3) is bound to notice systematic innacuracies, even when they only
>    happen to marginal applicants
> --
> David Canzi

I presume that you are suggesting that the second interviewer compile
the RESULTS of the first interviewers work and analyze them
statistically.  Mere knowledge of such a system will cause most first
interviewers (who wish to keep their job) to alter their behavior in
such a way as to damage their utility to the employer: they will
probably consciously select people on the basis of race, sex, etc., so
that their numbers look good (maybe even better than expected...).
Instead of making the hiring process color-blind, your system would
then make it color-conscious, with interviewers apportioning jobs by
informal quotas in order to preserve their position.  Is this an
improvement?

Also, you have avoided the problem of the second interviewer's
possible bias.  What if he were to simply not report discrepancies?
And if you are going to dismiss him if he fails to uncover prejudice,
might you not force him to "find" discrimination even if there is
none?

					David Rubin