[EM] small group methods

[Juho Laatu]

Hello Forest,

Here is one thought for a situation where
- election is non-contentious
- small group of voters know well each others

First of all, in non-contentious elections basic rating based methods 
work fine. For example in your second example of math hiring committee 
we can maybe trust that all voters sincerely want to pick the best 
applicant (that is not necessarily ones own favourite) and have no 
personal agenda to drive. (If some of the applicants are relatives or 
friends of some voter, then they could maybe indicate that, abstain 
from rating those candidates, and final results would be justified so 
that it would appear as if the voter would have given the applicant 
average rating of the other voters.)

The second example may thus be non-contentious enough to allow use of 
rating based methods. It is easier to estimate among a small group of 
voters if some particular election is non-contentious enough to allow 
use of rating based methods. (But of course there is also the risk of 
someone promoting use of a rating based method just because he knows 
that he himself has an interest to vote strategically (and hopes that 
others will not).)

My second point of voters knowing each others well opens up another 
possibility. The voters could also rate their trust on the other 
voters. Let's say that voter j gives rating Rij to candidate i. Other 
voters have rated their trust on voter j and his capability of 
evaluating the candidates. The sum of the trust points they have given 
is Tj. The final rating of candidate i will be sum of Rij * Tj (over 
all voters j). We thus get weighted ratings as a result.

(It is possible to set limits (min and max) to the given trust points 
(if you don't trust the voters to do that themselves; or to give each 
partner at least some weight). But if the election and partners are 
truly non-contentious one can of course just forget all such 
restrictions.)

This voting method might be practical e.g. when a small company of few 
partners votes about employee candidates. It may well be that some of 
the partners are more familiar with the area where they are hiring the 
new employee, and therefore other partners are happy to put some more 
weight on the vote of those partners that they believe to be experts in 
this particular area of activity within their company.

You already mentioned the possibility of repeated balloting when the 
number of voters is small. In the example above the partners could 
nicely arrange few rounds of balloting just to reduce the number of 
candidates down before the final election and to allow them to focus on 
the most interesting candidates a bit more. Let's say that they decide 
to reduce the number of employee candidates first down from 100 to 10. 
Then they would interview the 10 remaining candidates themselves, and 
discuss more about them. Then down to 3. And finally pick the best. The 
same above mentioned rating based method could be used in all 
elections. In principle the expertise of the partners would increase 
during the process and differences in values Tj should become smaller 
(or maybe the elections would become a bit more contentious towards the 
end ;-) ). Number of rounds and number of remaining candidates can be 
decided dynamically during the process - maybe with the help of voting 
methods to decide what to do next and how many candidates to keep :-).

Note that whatever ranking or rating based method we used above, the 
method should be able to list n best candidates (not just the winner). 
It is also possible to use some proportional methods instead of picking 
n most promising "singe winners" (e.g. to guarantee that all viewpoints 
will be sufficiently covered in the first round and corresponding 
potential winners will not be eliminated already at the first round). I 
leave that possibility for further study this time.

My feedback addressed mostly non-contentious elections (optimism? :-) 
). Maybe that is not as interesting to you than the contentious ones. I 
hope this mail stimulated some thoughts anyway. (In contentious 
elections requirements would be quite different, for example it could 
be necessary to limit the availability of information of the outcome of 
the earlier ballot rounds (both individual votes and compete results) 
=> ffs.)

Best Regards,
Juho


P.S. One more observation. In small groups it is easy to see if someone 
uses ratings in a bullet voting style (using only minimum and maximum 
points) if the ballots are published after the election (maybe to the 
small group only). This kind of behaviour would certainly lower the 
trust on that voter (and thereby the weight of his vote in future 
elections (and overall reputation as well)). Feasibility and possible 
balancing impact ffs. If the balancing effect is strong enough, one 
could even turn slightly contentious elections (of a small group of 
voters) into non-contentious elections.

P.P.S. In order to avoid problems when matters get contentious, it 
could be possible to make a rule that "contentious-proof" election 
methods will be used instead of ratings if some members of the small 
group want so. It is thus possible to build also mechanisms that allow 
dynamic selection of the used election method depending on how 
contentious the election is etc.


On Jul 31, 2005, at 06:01, Simmons, Forest wrote:

> I'm interested in hearing a variety of ideas for voting in small 
> groups where, for example, hands can be raised, repeated balloting is 
> not a problem, etc.
>
> One application I have in mind is text book selection.  At PCC all 
> sections of Calculus 251 must use the same text book.  The math 
> Subject Area Curriculum Committee, the "Math SACC," has the 
> responsibility for selecting the text.  The current practice is to 
> have a subcommittee narrow down the field, and then have the whole 
> Math SACC vote on the finalists.
>
> Another application is for the math hiring committee.  Currently this 
> committee narrows down the field from a hundred or so applicants to a 
> number between ten and twenty to interview.  After the interviews they 
> narrow the number down to something between three and six.  Then the 
> campus dean makes the final decision of who gets hired.
>
> I have other applications in mind, too, such as a group deciding which 
> tourist attraction to visit next when there are several to choose from 
> and only limited time, or a deciding which restaurant to go to, etc.
>
> Forest
> <winmail.dat>----
> Election-methods mailing list - see http://electorama.com/em for list 
> info