On Apr 11, 2013, at 6:11 PM, Ray Pelletier <rpelletier(_at_)isoc(_dot_)org>
wrote:
All
The IETF is concerned about diversity. As good engineers, we would like
to attempt to measure diversity while working on addressing and increasing
it. To that end, we are considering adding some possibly sensitive
questions to the registration process, for example, gender. Of course,
they need not be answered and would be clearly labeled as optional.
The IAOC would like to hear from the community on this proposal. It plans to
make a decision on its 18 April call in order to make the changes in time for
the
Berlin registration and will consider all input received by 17 April.
Seems like I'm in a minority among responders, but I think it's a bad idea.
First, what is suggested is not a survey. It's part of the registration form.
This makes it totally non-anonymous. These sensitive questions are going to be
linked to name and affiliation on the form. There is good reason why such
studies are always done through anonymous survey. People answer questions
differently when they're anonymous vs when they're not, skewing the results.
Alfred Kinsey would not have gotten the results he got in his sexuality surveys
without strict anonymity for the responders. Definitely not in the 1940s.
The other issue is that this will produce statistically invalid results because
of non-response bias. Non-responders may be positively or negatively correlated
with several groups the survey would like to measure.
Dave Crocker suggested getting an expert. I don't think that would help. Such
an expert would tell you that the questions you can ask depends on the group
you are asking. Questions that would be acceptable in one country, would be
inappropriate in another. In Israel people are likely to answer sensitive
questions truthfully, while in Germany concern for privacy might make these
questions seem too personal. So how do you fit such a questionnaire to a, well,
diverse group such as meeting attendants?
I think skewed surveys are a worse basis for planning policy than just using
common sense (yes, I know that's just another name for our biases). Surveys
lend a scientific aura to data that is effectively non-representative.
Yoav