Re: Voting experiment

> Jeff wrote:


>>>> On 10 Jul 2014, at 6:04 am, Charles McCathie Nevile  
>>>> <> wrote:
>>>>> My strawman proposal:
>>>>> The purpose of the experiment is to enable W3C Team to gather data on
>>>>> whether a different voting system to our current "Multiple
>>>>> Non-Transferable Vote" system would change the outcome of elections,  
>>>>> and in particular, in ways that might make elected groups more
>>>>> broadly representative of the voters.
>> At the AB discussion, we also discussed how long we should run this  
>> experiment for.  My recollection was 3 elections.  Is that your  
>> recommendation?

My recommendation is that the experiment be allowed to run for at least 3  
elections that are seriously contested. There are enough variations  
between elections as is that the results of the experiment would have to  
be extreme in order to justify acting based on a single outcome.

>>>>> In elections for the AB and TAG, we provide a ballot that offers two  
>>>>> ways to vote.
>>>>> 1. The current system
>>>>> 2. You can rank as few or as many candidates, plus the option "no  
>>>>> (other) candidate". as you want, in preference order.

A suggestion was made that instead of adding this explicitly, we  
implicitly do it in counting votes.

However my sense is that the common behaviour for STV systems is to simply  
let votes become "exhausted" - i.e. not giving further preferences is the  
same as saying "I don't care who gets there", which is different from the  
positive statement "I do not want these people to represent me". I.e.  
instead of a weak mandate, there is a clear statement explicitily limiting  
the mandate.

What would be done if this case arises is up to W3C - in any event, the  
idea is not that all this data would be public, for reasons already  
explained multiple times, but that W3C would at least have better data  
with which to judge the degree to which the current voting system  
represents the desires of the members who vote.

>>>>> 1 indicates your most preferred candidate. Giving two or more  
>>>>> candidates
>>>>> an equal rank is a rational statement, and results should be  
>>>>> calculated
>>>>> accordingly.
>>>>> A completed ballot for 3 seats with 6 candidates could be like:
>>>>> check         Candidate name        Preference
>>>>> up to 3                             order
>>>>> [ ]            Alice                   [1]
>>>>> [X]            Byron                   [2]
>>>>> [ ]            Charlie                 [ ]
>>>>> [ ]            Daniels                 [3]
>>>>> [X]            Elliott                 [4]
>>>>> [ ]            Franklin                [ ]
>>>>>                No (other) Candidate    [5]
>>>>> (In a real vote, the order of names should be randomised. Not that  
>>>>> we do
>>>>> that now).
>>>>> A vote for "No (other) candidate" [0] would be considered a vote for  
>>>>> a
>>>>> hypothetical alternative instead of a vote being "exhausted" (as  
>>>>> happens
>>>>> if all the candidates voted for by a single voter have been  
>>>>> determined as
>>>>> elected or not before the completion of counting). A candidate  
>>>>> beaten by
>>>>> the hypothetical alternative would not be considered elected.
>>>>> The results of this ranking can be used to asses the results we  
>>>>> would get by using simple "Single Transferable Vote" [1], "Schulze
>>>>> STV" [2].

>>>>> There are several ways to use votes as indicative of likely results
>>>>> from "Approval Voting" [3], although they are less reliable than the
>>>>> other information we would get from the survey.
>> Given that the Team needs to tabulate these results, it would be useful  
>> if there were available open source software to use for each of these  
>> schemes.  Do you know of any?  I assume that manual tabulation will be  
>> quite tedious.

Manual tabulation is usually done for simple STV in elections with tens of  
thousands of voters. In our context (perhaps 100 votes) it is relatively  
trivial to do manually, but I suspect there is software that will do it  

For Schulze STV, as I have noted many times in the discussion already,  
there is open source software available that will do the counting - see  
the end of the reference I already made.

In any event, we would need to get the individual ballots as input to the  
counting algorithm. Since this information is available from WBS, I don't  
foresee any complication there.

>>>>> [2]


Charles McCathie Nevile - web standards - CTO Office, Yandex         Find more at

Received on Wednesday, 23 July 2014 02:05:06 UTC