The intent behind requesting data from all of the small groups (Foundation staff, TC, Board) is to provide a view of the diversity of those groups compared to the community at large.
It is common for an organisation to demonstrate their progress in fostering inclusion/diversity, or lack thereof, by publishing these numbers. Essentially this demonstrates the extent to which glass ceilings exist in an organisation.
Similarly, if we survey our populations then we will have data to determine what the high priority areas are. Certainly, for the purposes of analysing the data on the general community, we will need to separate the data on small groups from the random samples.
I am intriuged by the idea of cross matching engagement levels with diversity data. That is an excellent idea! I will add a "please rate your level of engagement with or participation with the OpenStack community" type question. It will be fascinating to see if there are correlations.
On Tue, 8 Sep 2015 at 17:35 Eoghan Glynn firstname.lastname@example.org wrote:
My intent was to randomly sample and then offer an incentive. Thus we
would, hopefully get a very high response rate from a random group. Asking
the entire population with an incentive is definitely not a good plan as
Well, even offering an incentive at a later stage will compromise the
data somewhat (i.e. undo the effect of the initial randomization).
In terms of who we sample the answer is I think everyone.
In that case, I would recommend adding some questions to gauge their
level of active involvement in the community, so that we can slice and
dice the data accordingly.
e.g. it would be good to be able to hone in on the ATC subgroup of the
respondents, or the group working fulltime in the community in any capacity,
or the group working fulltime on OpenStack in more internally-facing roles
within their organization (e.g. operations, engineering/program management,
partner activity, marketing etc.)
We get a sample from each of the groupings, ATC would be one of them
obviously, but we also want to survey non-ATC community members (I believe
there are two classes of these).
It would be good to be able to collate those data from each of sub-
groupings alongside the totals.
It would be great if we could also request
all members of the PTL, TC, Board and Foundation staff groups to
I would advise against individually inviting any specific groups such
as PTLs, TC, board members and other community "nomenklatura", as again
that would simply serve to undo the randomization.
If the sample size is large enough and selected completely at random
it's likely to include some PTLs and/or TC members anyway. But including
all of the OpenStack nomenklatura in the sample would distort the data.
I would not want to send a message that ATCs are the only population worth
surveying. Our charter is much broader than that.
Fair enough, but I would strongly recommend that the survey is designed in
such a way to make it easy to drill down into the data specifically for the
more active sub-groups such as ATCs, or those involved in OpenStack fulltime
in some other capacity.
I, for one, wouldn't be that interested in building diversity among folks
who are only involved to the extent of voting in the board election every
year. Great if that larger group is diverse, but TBH it wouldn't be the
So it would be good to know where our baseline is specifically with regard
to the more active sub-groups with the foundation membership.
On Tue, 8 Sep 2015 1:48 AM Eoghan Glynn email@example.com wrote:
Mate, you should come to the meetings.
TBH neither of the alternating timeslots works with my schedule :(
A couple more thoughts though on the survey design ...
Up-thread someone suggested an incentive for participation.
I'm not an expert on sampling methodologies, but I suspect such an
incentive would actually amplify distortions due to self-selection.
Instead, it would be sounder from a methodological perspective to
randomly select a subset of the population to survey, rather than
relying on reactions to an incentive.
Which bring us to the question of what that "population" actually
is here, i.e. where are we most interested in measuring/promoting
I would lean towards the ATC community, since there's a requirement
to show some minimal concrete involvement (i.e. land a single patch).
Whereas the obvious alternative, the Foundation membership roll,
anecdotally includes many "paper" members with little substantive
involvement in the community.
Also I'd recommend removing the question on religion from the survey.
To my eyes, the question seems oddly out-of-place in a survey to be
undertaken within a technical community. The levels of adherence in
such a community may differ naturally from the overall population in
ways that make it difficult to reason over any data produced.
On Thu, 3 Sep 2015 7:11 PM Eoghan Glynn firstname.lastname@example.org wrote:
Interesting. Educational attainment as a proxy for merit ;)
Not a proxy for merit, more as a foundation for success in a
Education is one of the designated interest areas for the Diversity
hence it's inclusion. I'm happy to remove it of course if there is
consensus that it isn't appropriate or needed.
FWIW I'd recommend removing it.
Having said that, the question isn't whether something "should"
whole point of exclusionary practice is that it focuses on attributes
may not matter. What is important is what is used to discriminate,
analysing which of those are valid and invalid.
I'm concerned that we'd be getting into credibility-damaging territory
if we start talking about employers filtering candidates on the basis
of educational attainment as an exclusionary practice (with unfairness
and/or discrimination implied).
Is Education one of those attributes? Dunno. I've seen it used that
it certainly could be in our community (or in the hiring practices of
employers in the community).
When critically appraising hiring practices, IMO we need to carefully
distinguish between the innate characteristics of a person that do not
impact on job performance, and those acquired characteristics that can
and do so.
Just my $0.02 ...
On Thu, 3 Sep 2015 at 02:16 Eoghan Glynn email@example.com wrote:
I've added all the categories identified in the 3 phases
previously agreed, and altered the questions somewhat.
questions. I'm not keen to try to add any supplementary
Where I think we need to move beyond binary or simple data
I have stayed with free text entry.
I haven't yet written any introductory blurb about privacy
optional nature of the survey all the questions.
Similarly, I haven't yet addressed any issues around how
be targeted. I'm leaning towards a surveying a subset of
and trying to provide an incentive to participate (don't
one yet), so as to reduce self-selection bias. Anyone with
knowledge in this area please speak up.
I'm unlikely to make the next meeting, so I'm afraid I can
via email. We're running a little behind the original
to be able to engage the Foundation to commence the
the survey by the end of next week.
One area I always wonder about is English as a second
hamper efforts to get engaged in the community? I suspect so
to find solutions for further inclusion.
Can that be added if the goal of the survey is to identify
underrepresented people may be struggling?
It's an interesting question, but may be somewhat problematic
include in a survey.
For one thing, there are many in the community (who I've worked
who would fall into that category of
but would also have excellent proficiency in the language.
So simply measuring the number of non-native-speakers doesn't
tell us much in terms of hampered participation.
Also, it seems to cross the line between counting those with
characteristic (gender, orientation, race etc.) into counting
an (assumed) lack of mastery of a skill needed to thrive in the
Their proficiency can and will improve over time with sustained
the community can make allowances and level the playing field
say promoting co-presenters for design sessions or mandating
IRC as opposed to voice comms, but I would suspect that some
of baseline English fluency will remain long-term.
Good points, Eoghan. Why not phrase the question directly?
like "Does the fact that the OpenStack community communicates
primarily in English make it harder for you to participate?"
Sure, that's better - at least it only counts those who consider
themselves truly hampered by a language barrier.
Though thinking about it some more, and looking again at the latest
draft survey with the new question about educational attainment,
thinking that concentrating on innate personal characteristics
shouldn't matter in terms of participation) would serve us better
building diversity ... rather than straying into the area of
characteristics like having earned an under-grad/post-grad degree
(that do, and arguably should, matter).
Also, just to comment on the survey sampling, we got less
to our Women of OpenStack survey, so keep it in mind that we
meaningful data that you can act upon. We may need to dig
Foundation data and enhance those profiles instead, if the
ways to reach underrepresented groups."
On Wed, 2 Sep 2015 at 11:42 Johnston, Tamara <
The Diversity WG is actively working on many things,
forward with our Data Diversity Plan that includes
and how we’re currently collecting this information,
store this information, defining how to enable the core
report on this data, so on and so forth.
I support the stance the Foundation has taken in the past,
provide an open text field (and/or option to select
that enables a community member to, if they so choose,
identity. While we’re trying to better understand the
community we cannot limit the options they can choose
likely be perceived as personal questions (do you
minority). We can either choose to use an open text
say approach or take the hybrid approach that Facebook has
list 50+ identities but still have an open text field. I
with what the Foundation has been doing, as this will
members to decide if they want to share their sexual
not boxed into choosing X, Y, or Z.
Tamara Johnston | Cloud Portfolio | EMC Global
1-510-398-9114 | (E) firstname.lastname@example.org
From: Roland Chan email@example.com
Date: Tuesday, September 1, 2015 at 5:15 PM
To: Stefano Maffulli firstname.lastname@example.org, "
Subject: Re: [OpenStack Foundation] [Diversity] re:
The existing data is being handled by another sub-team on
WG. I'm certainly keen to see it, but getting it isn't my
Regarding the opt-out capability, my intent is that every
optional. The survey itself will require a one page intro
this and other privacy related issues.
On Wed, 2 Sep 2015 at 02:39 Stefano Maffulli <
On 09/01/2015 08:53 AM, Amy Marrich wrote:
I had sent this to a smaller section of the group but
the University of California asks the gender question
We may be able to get a hold of their survey as a
At the beginning of 2014, the OpenStack Foundations
its members to specify their gender. The intention was to
measuring that aspect of diversity in order to improve
gender issue is extremely new to society, there are lots
constant fluxes of differences among the non-binary
after long debate and research, to use an open text form
gender because that's the most flexible one. Any other
including the one from UC above, had criticism because
among scholars is not set.
You may have noticed that the form to subscribe to the
example asks gender offering 4 options:
- let me tell you
> open form
(I noticed now it's missing the very valuable 4th option
say", which I think may be useful to have even if the
Has anybody looked at the historic data about gender
 A summary of that conversation is on my blog