[EM] electowidget/robla

[Rob Lanphier]

On Wed, 2005-11-02 at 10:16 -0500, Warren Smith wrote:
> --suggestion:
> As both a test of your software, and as a public service and help to me, 
> could you run all the Debian elections using all (or at least a lot of) methods?  
> Hyperlinks to the debian elections at
>   http://math.temple.edu/~wds/crv/Debian2003.html
> (click years).  
> Then I will update this debian discussion page.

I'm hoping you can meet me halfway on this.  The data entry is  a bit
tedious, and not something that I have time to do right now.

Here's how to copy the template from the 2003 election:

STEP 1:

Convert the Debian ballots into JSON format.  This is the toughest part,
and something I needed to write a small convertion script to do.

Script for converting deb ballots (some editing and preprocessing
required):
http://electorama.com/2005/electowidget/util/debballot2jsonrange.txt

The script was written as a throwaway...something I didn't plan to give
anyone else, so I didn't write it for flexibility, and I don't plan on
enhancing it.

STEP 2:
Copy the source of the Debian election leader page:
http://wikitest.electorama.com/wiki/index?title=Election:2003_Debian_Project_Leader&action=edit

STEP 3:
Choose a new home for this script, and visit the URL.  Make up any name
you want...it's a wiki after all.  For example, you may choose the 2004
election:
http://wikitest.electorama.com/wiki/Election:2004_Debian_Project_Leader

STEP 4:
Click the "edit" tab

STEP 5:
Paste the configuration from step 2

STEP 6:
Replace the ballots in the configuration with ballots that you generated
in step 1.

> >> Second, I do not like negative numbers as votes.  They are confusing.
> >> Make it 0 to 10 not -5 to +5, say.
> >There seems to be some dispute on this.  I don't have a strong opinion
> >on the subject, but I'd like to hear more from others on it.
> 
> --One possible fix is to go with a GUI input style like mine in
>    http://math.temple.edu/~wds/crv/quickdemo/PresRadio.html
> but with the numbers 0-9 omitted from the display.  The point is that the voter
> then is inputting a "histogram" and there IS NO numerical range (or anyhow it
> is hidden) hence no way for voter to get confused.   Another idea is the voter 
> himself chooses the range to use, 0-10, 0-99, -5 to +5, or whatever.
> (From a menu, perhaps.  Keep track of the popularity of each menu choice in that case,
> please; I am interested in that data.)

Here's the same ballot, only using "radioarray" instead of "barscale" as
the ballot type:

http://wikitest.electorama.com/wiki/index?title=Election:Method_support_poll_%28radioarray%29/&action=edit

> >> I would like it to be totalled using lots of methods and see what happens.
> > The list: ... Definite Majority Choice
> 
> --this would seem to also require an "approval cutoff", which seems to require some
> kind of add-on to the vote-input gizmo.

The assumption made is that 1 or more is "approved", while 0 or less is
"not approved".

> >I am a little worried about load, so I haven't just piled every system
> >into every election.  Let me know the two or three others that would be
> >important to use.
> 
> --well, I very much doubt load is nontrivial, especially right now (perhaps if
> 100,000 users per day, then worry).  Seems to me there are 3 main purposes of electowidget
>  1. demo
>  2. public service
>  3. data gathering tool and source of experience about how well different methods work in reality
> For purpose (3) it is important to have all methods you can because that way people will
> be able to spot whatever weird phenomena arise, when they do, and we'll be able to see how
> often they arise.  I do not think purpose (2) and hence load will become important for some time
> into the future (development not yet ready for prime time).

Here's the problem.  A single page load takes a couple of seconds to
render already.  The more systems I put on, the longer a single page
takes to render.  There comes a point (something like 10 seconds) where
the server will refuse to do more processing.  It's a PHP setting that I
don't control.   So, even though aggregate load really isn't a problem
right now, individual page load is a problem.

> >> Fourth, I prefer it if we are allowed to offer "X" (no opinion, intentional blank)
> >> as a vote for something we know nothing about.   Your "0" vote
> >> can have the effect of supporting or denigrating a candidate (it depends on
> >> what the other votes are for the other candidates).
> 
> >How are you suggesting this get tallied?  Anything I add, I'd like to
> >have a way of mapping to all of the different systems Electowidget
> >supports.
> 
> --I don't know.  I am just saying, as a voter, I want this.  For example, I am
> asked to vote on about 100 voting methods, many of whch I have not heard of, and I don't
> want to take the time to think deeply about all 100 (say).  Right now the
> only way for me to express my ignorance is to give th unknown 0 or guesses.
> But if all methods are getting low or high scores, 0 will in fact have the unintended effect
> of making the unknown method look better than or worse than usual.
> 
> >"stealing" electowidget software
> 
> --good, I probably will steal it (later when it gets nicer).  However, it seems to rest
> on top of a lot of other systems (java, wikis), which might make it harder 
> to (or less attractive to) steal.  I might find it nicer to steal if there were simplified 
> version which rested on top of less stuff.

The reason why I rest on top of another system is so that I don't have
to write the user management code (account creation, email verification,
blah blah).  Granted, a wiki is a somewhat heavyweight host environment.
If there's a simpler PHP-based, open source piece of software that you'd
be willing to run, let me know, and I'll consider porting Electowidget
to use that system.  I'm currently already thinking about WordPress and
phpSurveyor, but I'm not wedded to those ideas.

Rob