Re: Fresh thread - 2023 Debian Project survey: Sustainability

[Joerg Jaspert]

On 17049 March 1977, Mathieu O'Neil wrote:

> • Survey responses are anonymous, IP and HTTP information are not
> logged, and all questions are optional. As it is still likely possible
> to determine who a respondent is based on their answers, results will
> only be distributed in aggregate form, in a way that does not allow
> de-anonymisation.

That "likely" is better worded as "for a great number of people
anonymity will not exist for anyone seeing the real answers".
Also, that aggregate form would need to somehow mangle the following 3
questions up pretty good, I think.

A03 formal status
A04 which 3 teams most involved
A05/A05a which other foss projects involved in

Those are enough to ditch about any anonymity I think, combine it
with B01c to make it easy enough for even an AI to find out who the
submitter is. I might be an extreme example, but for me it would be
enough to truthfully answer A04 (archive, new-maintainer, salsa) and
confirm it with A05a putting OFTC...

B01c only loses when someone works for a real big company.

--
bye, Joerg

[Joerg Jaspert]

On 17050 March 1977, Mathieu O'Neil wrote:

>>That "likely" is better worded as "for a great number of people
>>anonymity will not exist for anyone seeing the real answers".
>>Also, that aggregate form would need to somehow mangle the following 3
>>questions up pretty good, I think.

>>A03 formal status
>>A04 which 3 teams most involved
>>A05/A05a which other foss projects involved in

>>Those are enough to ditch about any anonymity I think, combine it
>>with B01c to make it easy enough for even an AI to find out who the
>>submitter is. I might be an extreme example, but for me it would be
>>enough to truthfully answer A04 (archive, new-maintainer, salsa) and
>>confirm it with A05a putting OFTC..."

> All response collected with LimeSurvey will be downloaded to a
> password-protected device in the University of Canberra and the data
> on LimeSurvey will be deleted after downloading. All the data analysis
> will be done locally, no cloud computing will be applied to prevent AI
> from obtaining the data. To better prevent reidentification, the raw
> data will not be shared to the Debian developers in our research team.
> Only aggregated results will be shared with the whole team for
> interpretation.

Are you an AI, or used one to write this? :)
Me mentioning AI was just a sidekick on how bad AI is, but the general
point is the *ease* of identifying people answering the question set and
that pretending it will be anonymous is a nice wish only.

>>B01c only loses when someone works for a real big company.

> The reason we kept this question from the 2016 Survey is to
> investigate how firms’ engagement in Debian has changed over time, but
> all the questions in the survey are voluntary, if you feel
> uncomfortable answering any question feel free to skip it.
> We could add a sentence in this question like “If you feel
> uncomfortable to answer it, feel free to skip it and move on to the
> next one”. Would that help?

I think the whole questions I mentioned should have something above that
the combination of the answers may make it really easy to identify the
person answering.

--
bye, Joerg