Re: list o' importants

[Roger Dingledine <arma@mit.edu>]

In message <Pine.SOL.3.96.980703152224.3833A-100000@kanto.cc.jyu.fi>, hvirtane@cc.jyu.fi writes:
>But we perhaps need a way to group somehow the questions and also group
>somehow the people who will answer the survey. 
>
>These two groupings might also be connected somehow. It might be the case
>that for one type of SEUL-user some group of questions is not so
>relevant. 
>
>How would it sound to start asking as a first categorizing of questions
>the general uses/tasks of computers/OSs of the user (to make the uses as
>the base of categorizing)? It might be more useful (and easier to get
>the answers) to know what the people are doing with the computer than to

Ah-ha! This is what I was trying to think of before. That way the
introduction page can be "pick which one you use your computer for",
and then it will provide the appropriate survey, and that survey will have
both a "personal information" section and an actual survey section. That
works, except...what about people who use their computer for more than one
thing? (More than two? More than five?) Maybe "check all uses that apply,
but you need to specify exactly one as the primary use"?

>know exactly who the people are? (For example I don't like the marketing
>strategy of some companies, who want to know all kinds of personal about
>me, before they start telling me, what kind of products they can offer.) 

It's true. I was thinking about gathering personal information (eg, name)
for indexing and archiving purposes -- so a user can pull up his survey
response later and peruse or modify it. And I was thinking about keeping the
email address around in case we needed to contact the person.

But I'm beginning to think that's stupid. It will make people paranoid (and
rightly so), and we won't actually ever use it, so we shouldn't collect it.
We will not allow a user to modify a survey that has been submitted, because
that opens up another can of worms for both authentication (how do we know
it's the right person changing it?) and privacy (if it isn't, why do they
get to look at that person's answers?)

But I still think we should collect data for age, income, etc, because those
are statistically relevant. And somewhere in there I definitely want to ask
about current operating system preferences, because I think that's
extremely relevant.

>More categories to be added by the people first surveyed.
>
>(Somehow it so clear and simple that any OS should be cheap, reliable,
>easy to use, fastdeveloping, powerful, full of new possibilities and
>advancements of technology that it isn't necessary to ask, if the
>would-be-users want these qualities or not... Would it generate
>much helpful knowledge, if we knew, which of these are the most important
>for which type of user? Or is it more useful first to know, which kind of
>users exist and how many?) 

I think it will be very difficult for us to figure out "which kind of users
exist and how many" without some expensive survey techniques to make sure
we get a fair sample. There will be parts that are hand-wavy because this
is not a professionally administered survey. 

Figuring out what features are most important to which types of users
is the goal of the survey, if only because I think it's the only thing we
can reasonably expect to figure out.

>My general idea is to find out somehow, what is the current situation of
>computing (with microcomputers), find out next what are the general 
>problems, which people wanted currently to get solved, find out next if
>linux could be a part of solutions of current problems.  
>
>This leads to the next step of categorizing: which kind of problems and
>priorities of solving of those each type of user has in each category. 
>If the categories are cleverly planned the survey doesn't need to be so
>big, but to find different kind of users to invent the categories we maybe
>needed to be quite clever?

Wow. So what we're proposing right now is to come up with perhaps 10 or 15
different "types" of users (divided based on why they use a computer), and
then take our list of issues we care about, and figure out the correct
order for presenting each of those issues to each of those user types.

That's a tough assignment. :) Is it worth it?

The downsides are that it's a lot of work, and that it might confuse people
when the surveys don't look the same between user types. I think it's a
major win in terms of being able to compare survey answers from the same
user type, though. The interpretation and comparison side of things will
be much easier.

>(It really might be the case, that computers are now mature enough so that
>the problem of speed or power for some tasks is already solved... Like
>cars of sixties were already fast enough for family use... 286 pc with dos
>and wp 5.1 is as good or even better for writing a novel as 300 mHz
>Pentium II fitted with linux and applix 4.3.7! ) 

Are any of our current questions designed to be able to notice if this is
the case? It's a neat thing to be able to conclude, but I can't imagine
any combination of answers to our current survey questions that would
let us notice this. Hmmm...

--Roger