Re: [Ag Meinungsfindungstool] [Ag-buergerbeteiligung] Score Voting

["Christoph \"Pluto\" Puppe" <piraten AT stderr.de>]

Am 13.05.2012 01:19 schrieb "Clay Shentrup" <clay AT electology.org>:
>
> On Sat, May 12, 2012 at 4:01 PM, Christoph "Pluto" Puppe <piraten AT stderr.de> wrote:
>>
>> > I don't like hard quorums like that. We used to have a hard quorum for doing "safe" average-based Score Voting (instead of sum-based). But we improved it by stealing an idea from IMDB.com, which was actually quite mathematically savvy.
>> > http://ScoreVoting.net/BetterQuorum.html
>>
>> Failed to understand this :) Do you say, the no opinion is not counted?
>
>
> I'll respond to the rest later, but this part I can answer quickly now.
>
> Say you just use total sum of scores. Then what can happen is that some very good candidates can lose simply because they are not widely known. We saw an effect like this with the Time magazine "Top 100 people" poll a few years back. Craig Venter, the scientific luminary behind the Human Genome Project, had nowhere near the number of points of celebrities like Lady Gaga or whom ever.

Ok, i dont like venter, patents on genes ... but i do love lady gaga, the sole representative artisan of a generation! And he impersonates, everything, the culture anthropologists will pin on this time: post-gender, change, direct contact, post-hedonism ... ok im drifting here. Its sunday :)

> But if you looked at average points, Venter and others like him were at the top. Because the people who did know enough about them to give them a rating were those who were more educated. But if you go only on averages, it's possible for some relatively unknown maniac to win. His three supporters give him a "perfect 10" average. You get the next Hitler perhaps. :(
>
> So how can we get the best of both worlds? Well, the system discussed on that link does some very sophisticated math to arrive at the answer. But it turns out that, once you simplify it, you get a very simple result. You can just use averages, but give each candidate a bunch of 0's to start with. So for instance, say you give ever candidate 20 0's. Then he gets a score of 8. So his average score will be 8/21 = 0.38. The 21 is 20 zeroes plus the 1 score he actually got. At first, this is a severe handicap. But say he gets 30 real score, averaging to 8 points. Then his score will be 4.8. Those 20 zeroes are still a strong handicap. But then say he gets 200 real scores, averaging an 8. Then his average score will be 7.27. The zeroes eventually have a very tiny effect, once a statistically significant number of votes come in.

Ok, now i got it! Put that way its easy :)

Actually we had the discussion last week how to handle new entries in reputation voting. So a new person has no reputation. Old hands might choose to filter based on high reputation, result: no fresh ideas reach the leathery ears ...

To prevent this, new accounts geht the highest possible reputation. (In a system with finite scores, this unlimited reputation thing of the tool, was not known then). As repuatation should be points/votes, an average and the starting reputation should habe some resilency ... we thought to have 5 ghosts account vote on each new account. Pretty much the same idea.

Select a sensible start, some weight and then let loose.

> 7.27 could allow him to beat someone who has an average of 7 points from 100,000 voters. The idea is that once you have a statistically significant sample size, you don't need to keep handicapping. The zeroes gradually play a smaller and smaller role. But note I said gradually. It's a "soft" quorum rather than a "hard" quorum that happens all at once at some arbitrary threshold.

Like that. Ok, i think the software has to change ... damn :(

> So the formula is to pick the number of artificial 0's you want to give, and then just dived the total points by the number of voters plus the number of zeroes. Very simple.
>
> For practical purposes, it's probably more than sufficient to just use total scores. But for groups like the Piratenpartei, which have a lot of very smart people with good math skills, this extra complexity could be worth the cost. You don't want Craig Venter to lose to Lady Gaga just because more people know who she is. :)

See the start of the mail :)