[lug] AI survey.

Tue Nov 21 14:45:42 MST 2017

Just to add a bit more nuance, Jeff's suggestion "that middle values be
used when a scale doesn't correspond to reality" simply raises the issue of
well-designed scales.  I entirely agree that the "left <-> right" scale
really isn't a good model of political belief.  Even adding another
dimension, like the libertarians' model that uses an 'authoritarian'
spectrum along the vertical axis still seems like a pretty large
oversimplification to me.

But a middle value still isn't a good representation of Jeff's ideas.
Putting a checkbox in for "none of the above" is at least a way to identify
whether a question might be problematic, but then we introduce the
possibility of providing too many choices and decision fatigue.  A better
approach would be to have a way of measuring someone's confidence in their
answer - data on the time it takes for someone to choose an answer, or if a
later question causes someone to go back and revise a previous answer...
but even then very subtle differences in the phrasing of a question can
result in dramatic differences in results.  One of the most notable
examples of this (dangerously keeping with the theme of political polls) is
the question "Do you approve or disapprove of the way the president is
handling his job?"  This phrasing has been used as the standard for
evaluating the "presidential approval rating," and yet it's obviously not
measuring something people would describe as someone's approval of the
president.  Handling his job?  Hmmm.

I'd be inclined to just give everyone text fields for everything and
analyze word/phrase uses, but that just shows my own biases towards data...

And yeah, "chosen field" made me stop and think a while.  I've recently
been doing some devops-y things some of the time, and helping out with the
infant hearing screening program at Boulder community hospital as a
volunteer some of the time, and tinkering with Rust.  My "chosen field"
might be better answered with "yes" than a list of multiple-choice
options.  (This makes applying for jobs much harder than it really should
be.  I'm currently looking for employment.)

On Sat, Nov 18, 2017 at 11:33 AM, Jeffrey S. Haemer <
jeffrey.haemer at gmail.com> wrote:

> I'll disagree with Chris, just for the reason he points out: uncertainty,
> and add that I choose middle values when I think a scale doesn't correspond
> to reality, or assumes a view of the world that isn't mine, as with the
> "left <-> right" political scale on the AI questionnaire.  If I'm lucky,
> there's a checkbox that says, "None of the above," or "Decline to answer,"
> or "Gosh, I'm not sure," but there seldom is.
>
> Plus, for old folks like me, the "chosen field" question gets tougher.
> Would Jimmy Stewart have put "movie star" or "Air Force general"? Does
> Chris put "linguistics" or "programming"? What would her dad put?
>
> On the other hand, I'd give even odds that your daughter didn't make up
> the questionnaire, and has little or no input on it. We all just answer as
> best we can and she -- or some AI program -- looks for patterns.
>
>
> On Fri, Nov 17, 2017 at 4:10 PM, Chris Riddoch <riddochc at gmail.com> wrote:
>
>> Thanks for posting this, Dave.  I like the questions chosen there, and am
>> interested in seeing the results of the project when it's ready!
>>
>> As an aside (because the survey is already underway) I'd recommend future
>> surveys omit a 'middle' option from some of the range-based questions by
>> using an even number of ranged options.  Middle options can be too tempting
>> an answer for lots of reasons, like being conflated with uncertainty.  I
>> don't know how I'd go about trying to figure this out (much less make the
>> answers comparable), but I'd be very interested in knowing why someone's
>> chosen particular answers for some of these questions.
>>
>> Cognitive science is a really fun field.
>>
>> On Thu, Nov 16, 2017 at 8:07 AM, David L. Anselmi <anselmi at anselmi.us>
>> wrote:
>>
>>> Hi BLUG,
>>>
>>> My daughter is conducting a survey on artificial intelligence to collect
>>> data for her undergrad data sciences class project. Would you be so kind as
>>> to participate?
>>>
>>> It takes 5 min or less and you'll help shift the demographics away from
>>> the people she knows on social media. :-)
>>>
>>> https://goo.gl/forms/8H9sUHrjoUZDbtCT2
>>>
>>> Thanks!
>>> Dave
>>> _______________________________________________
>>> Web Page:  http://lug.boulder.co.us
>>> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
>>> Join us on IRC: irc.hackingsociety.org port=6667 channel=#hackingsociety
>>
>>
>>
>>
>> --
>> Chris Riddoch
>> http://www.syntacticsugar.org/
>>
>> _______________________________________________
>> Web Page:  http://lug.boulder.co.us
>> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
>> Join us on IRC: irc.hackingsociety.org port=6667 channel=#hackingsociety
>>
>
>
>
> --
> Jeffrey Haemer <jeffrey.haemer at gmail.com>
> 720-837-8908 <(720)%20837-8908> [cell], @goyishekop [twitter]
> *פרייהייט? דאס איז יאַנג דינען וואָרט!*
>
> _______________________________________________
> Web Page:  http://lug.boulder.co.us
> Mailing List: http://lists.lug.boulder.co.us/mailman/listinfo/lug
> Join us on IRC: irc.hackingsociety.org port=6667 channel=#hackingsociety
>

-- 
Chris Riddoch
http://www.syntacticsugar.org/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lug.boulder.co.us/pipermail/lug/attachments/20171121/f1d42c8b/attachment.html>