[EM] democratic community, the web, implicit/explicit instant proxy

[raphfrk at netscape.net]

 From: robla at robla.net
 > > And a meta note, I guess the "election methods" list isn't just for that 
 > > little issue of "voting" and "elections" anymore. Sometimes we branch out 
 > > into the whole field of the organization of democratic societies.
 > 
 > I think what I'm proposing is for voting and elections, but the target
 > structure is not a typical government leader.
 
 It seems that what you want is just a filtering system. However, 
 maybe that could also be considered a concensus system.
 
 It could become an election system if it was combined with a 
 voting method. If it was really good at achieving concensus the vote
 could require a supermajority.
 
 For example, the method would be:
 
 Anyone can submit a proposal for a change as an article. Also, they can
 submit an article about a specific change proposal.
 
 The filtering system makes sure that the "good" articles become well known.
 
 A change requires a simple yes/no type vote. Anyone who abstains is
 considered a no vote. This means that to implement a change, your idea
 must be supported by the majority and become well known. Depending on
 how good the system is at achieving concensus, a supermajority could 
 even be required.
 
 Also, two proposals could be linked such that they together are 
 supported by the majority.
 
 For example, the utility of 2 changed A and B could be different 
 for each group 1, 2 and 3
 
 Proposal A
 1: +10
 2: -10
 3: -5
 
 Proposal B
 1: -5
 2: -10
 3: +10
 
 Both would only have support of one group in 3. However,
 
 Proposal A+B
 1: +5
 2: -20
 3: +5
 
 would have support of 2/3 and thus be passed.
 
 From a social utility point of view this is actually bad. (Total 
 utility drops). This could be solved by requiring a supermajority.
 
 Arguably, one of the benefits of a really good concensus builder is 
 that it lets you require a higher supermajority to get things passed,
 without it causing deadlock. This prevents more things that are 
 on balance a bad idea from being implemented.
 
 > 
 > > Problems:
 > > Too much stuff
 > > Too much mediocre stuff
 > > Too much good stuff getting lost
 > > Different users desire to see different content
 
 <filtering>
 
 > > Web user interface is often slow and limited
 
 well that is a tech question and is getting better. Also, there is a 
 reasonable amount of custom code that is possible. What kind of interface
 were you thinking of ?
 
 > Those are problems for readers. However, if you're a writer who isn't
 > yet one of the "cool kids", you've got those problems, plus:
 > * Small incentive for people to read your stuff
 
 So you need to create an incentive. Clearly being a popular writer
 is incentive for alot of people. Perhaps adding a system where 
 reviewers are also given a rating.
 
 A user could have a high writer "score" and/or a high reviewer "score".
 
 In theory, there could be lots of ratings. However, to keep it simple
 lets say there is just 2. 
 
 Accuracy -- This is your opinion of the accuracy of another reviewer's 
 ratings
 
 Interesting - This is your opinion of how interesting an article is
 
 The system could work as follows:
 
 To determine the interesting score of an article, the following process
 is followed by the system
 
 1) Check if you have rated it
 If so, return
 Accuracy: 1.0 (or maybe allow you to set it lower)
 Rating: The rating you gave it
 
 2) Pick a random user who you have an accuracy rating for AND
 who has rated the article
 
 Poll their opinion of the article
 Update your rating and accuracy estimate based on
 Your opinion of their accuracy
 Their opinion of the accuracy of their rating
 Their stated rating
 
 3) Return to 2 until you have reasonable accuracy or have tried
 a threshold of users.
 
 Every time you rate an article, it checks who else has rated the 
 article. Your opinion of the other user's review skill is 
 modified based on how closely they rated the article compared
 to you. The lower they rate the accuracy of a given rating,
 the less of an effect it has on their score.
 
 This means that if someone rated unrated articles quickly, 
 they will be checked against everyone who rates the article
 later. This means that early and accurate ratings are 
 encouraged.
 
 The system would go on automatically. You will compute ratings
 based on your opinion of other reviewer's ability. These will
 then propagate based on other users' opinion of your rating
 ability. Also, your accuracy ratings will be updated. 
 Everyone will then get a personal rating for each article. 
 
 The more articles they rate the more consistant the score they 
 get for the articles. This should balance out. If users are 
 getting bad accuracy ratings, they can improve it by rating
 more articles. This improves everyone's ratings quality.
 
 > * Too much riding on SENSATIONAL headline
 > * No motivation for readers to step outside their comfort zone if what
 > you present doesn't reinforce their world view.
 
 If there is a decay on older ratings, then perhaps a person could 
 move from their comfort zone. For example, they might read an
 "interesting" article that moves their comfort zone slightly. However,
 it wouldn't be a massive change, so won't trigger the "this is an 
 'enemy' viewpoint, must switch off reason and switch to 'tribe' mode" 
 effect.
 
 > * No way of knowing where you really stand, and who you should lobby
 > next
 
 I guess you could include a report of who tends to rate your 
 writings lowly. Also, quite possibly, the users would split into 
 groups. A group wouldn't have a hard edge and would consist of
 writers and reviewers who tend to rate each other's accuracy and
 interestingness highly.
 
 Maybe this could be displayed graphically. You could even see
 "nearby" users as potential people to convince to agree with your
 opinion.
 
 > > So, in summary, I think what I want is:
 > > Variable Share Instant Proxy with Direct option - Representatives 
 > > get as much share as they have constituency. Anyone can vote directly on 
 > > anything if they choose.
 > > Implicit Proxy - Extracted from various user actions, up-rating 
 > > comments, recommending diaries, etc.
 > > Explicit Proxy, Explicit acclaim - when the user takes the time to 
 > > say "I like this other user!" that should be overriding or heavily 
 > > weighted.
 > > Randomized Presentation - show a mix of directly desired, 
 > > indirectly desired, popular acclaim and random content. Promote 
 > > recommended content through the network of the recommender out to the 
 > > wider community.
 > > 
 > > And yes, I may just get busy and code it myself, but writing a brand new 
 > > wiki/blog/community infrastructure from the ground up (because PHP sucks 
 > > and perl is cumbersome at large scale) will take a while.
 > 
 > Yeah, whatever system would need to be bolted onto the side of some
 > existing system, I think, in order to be viable.
 
 I think another big issue is that something like this is actually annoying
 until you hit the scaling issues. In a forum with say 20 users, you don't
 actually need it. However, by the time a forum hits the "big time", the
 system is already in place. Changing it to a system that scales better
 is harder at that point.
 
 I think that a possible method would be to find some community that is 
 suffering from the scaling problem and offer them a solution. 
 
 On forums, there is sometimes a rating system. However, they seem to 
 just be a +1 button you can hit as people abuse it. In the suggestion
 above, that should be less effective as badly rating someone will just
 mean others will lower their accuracy rating of you. Also, in a 
 conversation type format, blanking out posts results in the "flow" 
 being lost. (though I guess that is exactly what a proxy system is 
 designed to solve).
 
 I wonder if a way to do it would be to have good/neutral/bad for 
 articles/posts by default. However, a post that achieves a strong
 good score might switch to the users being able to choose 
 very good/good/neutral. The users are always given 3 choices
 centred on the current score.
   
________________________________________________________________________
Check Out the new free AIM(R) Mail -- 2 GB of storage and industry-leading spam and email virus protection.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20060830/60e45a57/attachment-0003.htm>