Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!rice!brazos.rice.edu!bbc From: bbc@nysa.rice.edu (Benjamin Chase) Newsgroups: news.groups Subject: Re: Proposal for changes to the newsgroup creation guidlines. Message-ID: Date: 17 Oct 89 07:53:21 GMT References: <14718.2538b6f4@kuhub.cc.ukans.edu> Sender: root@rice.edu Reply-To: Benjamin Chase Distribution: na Organization: CPRC, Rice University Lines: 147 In-reply-to: sloane@kuhub.cc.ukans.edu's message of 15 Oct 89 21:57:23 GMT Bob Sloane presents a proposal: >5) Votes should be cast [using wildcards] >7) Counting the votes may be done as follows... >Please keep in mind that this is just a proposal. I welcome any comment >anyone has to make. Well, you asked for comments, and mine is that your wildcards were too complex. I'll admit your scheme is concise, but IMHO is too confusing for many users. (Just as "IMHO" and similar abbreviations are concise but confusing to USENET novices.) A user-interface expert (such as the guy with whom I share an office) would say that your wildcard specifications are "too unixy". Perhaps you've been using Unix and its syntax for regular expressions for so long that you don't even see this possibility? Certainly it is pleasing to some people, such as those who like mathematical puzzles, to produce the shortest ballot possible. But if a person is only voting once, how much economy is really derived from this, and how sure can a voter be that she's actually said what she meant? From a personal letter that Bob sent me: > How do I vote against any group in the sci hierarchy and for any group > in the rec hierarchy? I assume the call for votes would have to list all > the possible group names. What if I want to vote for some other name? > It seems simpler to me to say: > > sci.* no > rec.* yes > > rather than give a list of 10 possibilities. In the current sci.aquaria > debate, the suggestions that I remember were: > > [list deleted, but presented below] > > Suppose I just want a group created, and don't care about the name. Do I > list all of these? To the last question, I say "yes". In the modern world of software, I wouldn't type in the list of candidates, but just clip it out of one letter, drop it in another, and edit it a little. ---------------------------------------- Let's compare two other schemes. First, the ballot elided above, written in Alien Wells' style: sci.aquaria no sci.bio.aquaria no sci.aquarium no sci.bio.aquarium no sci.aquariums no sci.bio.aquariums no rec.aquaria yes rec.aquarium yes rec.aquariums yes rec.pets.aquaria yes rec.pets.aquarium yes rec.pets.aquariums yes rec.pets.fish yes This ballot was not too tough to complete, and very obvious in meaning. Alien has already outlined ways in which ballots of this type would be tabulated. I leave the nits, complaints, and implementation details to him. ---------------------------------------- Next, I offer the same ballot, cast according to a preferential scheme. In this system, the choices are reordered by the voter in order of her preference, most preferred first. This scheme would be an excellent choice for the current quandary over alt.aquaria's new name, since it allows many names to be considered in a single vote, and allows the voters to rank them in that single vote. One assumption is that the voter actually has fine-grained opinions about each choice. Note that an extra choice, "none-of-the-below", is added to the list of candidates to indicate the voter's distaste for all the names that follow it on the list. The ordering of the candidates that fall below "none-of-the-below" is unimportant. rec.aquaria rec.aquarium rec.aquariums rec.pets.aquaria rec.pets.fish rec.pets.aquarium rec.pets.aquariums none-of-the-below sci.aquaria sci.bio.aquaria sci.aquarium sci.bio.aquarium sci.aquariums sci.bio.aquariums Now, I haven't worked out all the details of this, but... After collecting the preferential ballots, tally them according to their first choice. Next, iteratively throw out the lowest vote-winner (that is, the "loser"), redistributing its votes to the other candidates, according to the next highest choices on the ballots of that loser. Special treatment is needed for "none-of-the-below" (or whatever you want to call it. I chose that name to make sense from the voter's point of view.). During the iteration mentioned above, this candidate should not have its votes redistributed to other candidates, even if it is (currently) the lowest vote-getter. It can never "lose". Instead, "none-of-the-below" should be passed over, allowing it to accumulate votes, and the next lowest vote-getter should be picked instead. The total votes received by "none-of-the-below" represents the number of voters who consider all the remaining candidates unacceptable. Thus, if rec.aquaria where the loser in the first round, all the votes for rec.aquaria would be redistributed. The above ballot would be counted towards the total for rec.aquarium. If that candidate were to lose later in the tallying, the above ballot would be counted for rec.aquariums. If all the choices that appear above "none-of-the-below" lose, the above ballot would be counted as a veto on the creation of the new group. The process stops when one candidate attains a particular victory condition. If the winning candidate is "none-of-the-below", no group is created. Possible victory conditions might be: 1) majority of votes cast 2) majority, coupled with a certain margin over "none-of-the-above" 3) a certain percentage margin over all the other candidates 4) ? I leave the choice of the victory conditions to the more experienced members of news.groups. Results of voting should include the order in which losers were thrown out, and the number of votes gained by all candidates remaining when the winner achieved victory. To be thorough, one could also report where each person's vote ended up, allowing each voter to verify that their ballot had been followed correctly. Note that this voting scheme allows write-ins, and that reporting the losers in the final results gives the write-ins visibility. If no name wins the vote, perhaps the write-ins could be required choices on any subsequent vote for the proposed group. So, does anyone have opinions on _this_ scheme? -- Ben Chase , Rice University, Houston, Texas