[EM] Does Range need an abstention/participation tally?

[Kristofer Munsterhjelm]

On 02/22/2012 11:15 PM, Jameson Quinn wrote:
> I'm working on sketching out data structures so that Helios Voting
> <https://vote.heliosvoting.org/>, an online, open-source,
> cryptographically-verifiable voting system, can use advanced voting
> procedures such as Range, Majority Judgment, and SODA. (Condorcet is a
> significantly harder problem but probably doable, and IRV is essentially
> impossible).
>
> My question is: for the Range voting structures, is it acceptable to
> just keep one tally (total score) for each candidate, or do you also
> need a tally of number of voters who rated/didn't rate a candidate? The
> latter would be used for average-based schemes; so this question is
> equivalent to asking, are such schemes important enough to be worth
> making the data structures more complex? Since I'm the one signing up
> for the programming work here, I'd appreciate it if answers that ask me
> to do more work have a reasoning and a strength (ie, "I'd kinda prefer
> it" versus "I think it is absolutely necessary").

Average Range would be nice but I don't think it is essential. If you 
did implement it, I think you should also have a soft quorum parameter 
so as to avoid the case where an unknown get ten max-ratings from his 
close friends and so wins.

The most reasonable such, I think, would be to simply add a certain 
number of min-rating votes to every candidate. I don't know what that 
certain number would be, though.


Incidentally: perhaps you could get around signature-type attacks in 
Condorcet by having people not give a full ranking, but an answer to "is 
X better than Y" for a randomly picked pair {X,Y}. Sample the pairwise 
matrix instead of asking for all of it. Then vote-buyers wouldn't be 
able to say "rank in this specific order and I'll reward you". I've been 
thinking about that in the context of simple cryptographic voting 
methods (VAV, ThreeBallot, Twin), so I don't know if it's applicable to 
computerized crypto methods, but I thought I should mention it :-)