Talk:Score voting
move to "range voting"
Hi there. I hope that I won't offend anyone by this move. I just don't think that it makes any sense to call this method "cardinal ratings". My main reason is that "cardinal ratings" describes a ballot type rather than a tally method. A secondary reason is that the term "cardinal ratings" is somewhat redundant in itself.
If people really don't like the name "range voting", I'd suggest "ratings summation", or "ratings sum".
I suppose that "average rating" works as well, but it raises a question: what is the impact of a ballot that doesn't rank a particular candidate? If the ballot counts as the lowest possible score (e.g. 0 on a 0-100 scale), then "ratings summation" is the most appropriate name. If the ballot doesn't affect the candidate's score at all, then "average rating" is the most appropriate name. For political election purposes, I think that a candidate not ranked on a given ballot should be assigned the lowest score, so that candidates who are unknown to most voters are unlikely to win. Hence, I favor the term "ratings summation" rather than "average rating".
Am I making any sense?
My best, James Green-Armytage 04:39, 19 May 2005 (PDT)
- I don't object. This is under "range voting" on wikipedia as well. I suppose "CR" was popularized by Mike Ossipoff. I'm curious to know who coined "range voting," though. I first remember seeing it in Warren Smith's paper (which was very opinionated and contained a number of errors). Kevin Venzke 11:38, 20 May 2005 (PDT)
- Good question; I don't know. Maybe a good topic to bring up on the list. I was under the impression that it had a fairly well-established usage, but I could be wrong. Annoyingly, most of the google hits I get are clones of the wikipedia article (with this one at the top), and none of them seem to cite a seminal paper. Even if "range voting" doesn't have a good usage history, it's still probably better than the redundant "cardinal ratings", but perhaps "ratings summation" would be more descriptive? -- James Green-Armytage 10:12, 21 May 2005 (PDT)
Condorcet Criterion violation
- Comment on the Condorcet Criterion violation example made 2008-08-05; reformatted by User:RobLa on 2020-05-04
If 51 voters vote A100>B90 vs. 49 voters with B100>A0, then isn't it possible that B-preference voters have an incentive under Range Voting to bullet vote for B instead of rating A truthfully? In that case, Condorcet correctly avoids giving that incentive. --Araucaria 17:40, 5 August 2008 (UTC)
Alternate title suggestion: Score Voting
The rangevoting.org site has recently been proposing Score Voting as an alternative term. --Araucaria 23:21, 13 August 2009 (UTC)
Example is flawed
IMO, the example of converting ranking to scores is flawed, and I assume this is inherited from the Wikipedia source. The example with scores from 1 to 4 is essentially the same as Borda, and doesn't represent what voters might do in an actual Score Voting election.
One might reasonably expect that Memphis voters would vote 100% Memphis, 90% Nashville, 30% Chattanooga, 0% Knoxville, which would have a much different score result, though probably Nashville would still win. --Araucaria 23:21, 13 August 2009 (UTC)
Strategy
Range Voting could be interpreted as a method to approximate the centroid of a data set by finding the choice with the smallest sum of squared distances to each voter. However, since an individual voter has no way of measuring absolute distances, the best they can do is to compare their relative distance from every candidate, giving their closest candidate the highest score and the most distant candidate the lowest score. The question then is, what function of relative distance best approximates that least sum of squares property? --Araucaria 23:21, 13 August 2009 (UTC)
Normalization and rational voters
User:Dr. Edmonds, while I'm not going to alter your edit (https://electowiki.org/w/index.php?title=Score_voting&type=revision&diff=10865&oldid=10864), I do want to point out that some Score advocates don't actually believe all or even most voters will normalize in practice, so it is better for the sake of wiki neutrality to phrase your belief in normalization as what most people think rather than the absolute truth. See (https://www.reddit.com/r/EndFPTP/comments/7v2qa2/comment/dtrr6z7), (https://www.reddit.com/r/EndFPTP/comments/7v2qa2/comment/dtsa8hn), and (https://www.reddit.com/r/EndFPTP/comments/7t0uzp/_/dtiyw40). BetterVotingAdvocacy (talk) 20:57, 4 May 2020 (UTC)
- I have never met a single score advocate who does not think voters would normalize. By normalize we mean give their most preferred the max score and their least preferred the min score. It would likely be part of the educational material around the system and would be advised for clear strategic reasons. Weather they do this or not, my edit was to write it in a way to make it clear that these results are contingent on such weird behaviour. I could be totally wrong on this and it might be the case that a significant number of voters act stupidly. If this is the case I might stop advocating for score and switch to approval. I intend to ask such a question at the event tomorrow. https://www.electionscience.org/presidential-election-analysis/ This is likely the best way to know for sure.--Dr. Edmonds (talk) 01:46, 5 May 2020 (UTC)
- Wouldn't you consider the common "Score allows a consensus candidate to beat a polarizing majority-preferred candidate" argument to be implying that voters won't necessarily normalize in a two-candidate election, and that indeed, this is a feature of Score, not a flaw? To add, my own argument that voters won't necessarily normalize is that for voters who are wavering between voting and not voting in a particular race, there is nothing "irrational" about them going in the middle and essentially casting a partial vote. BetterVotingAdvocacy (talk) 01:56, 5 May 2020 (UTC)
- I do not think the consensus candidate argument applies or is even a definable statement in a two candidate race. I think in summary it is possible that voters do vote in lots of crazy ways that ultimately hurt them. Steven Brams says this is why he prefers Approval to Score. he does not trust them to vote properly. It is ultimately an empirical question. Talking about things like grading in the Olympics is not a reasonable analogy. We need examples of actual high stakes political races. I would think the talk tomorrow should provide some empirical evidence. --Dr. Edmonds (talk) 03:17, 5 May 2020 (UTC)
- BetterVotingAdvocacy It seems the problem is not investigated. I am going to push them to do so in their upcoming paper. I did come across this other paper which as two important results. The first being that when using a scale with negatives voters tend not to give unknown candidates a negative score as they should. This means [0,x] scales should be used. The second result is that different x values produce consistent scoring. This is good to see. --Dr. Edmonds (talk) 18:44, 7 May 2020 (UTC)
- BetterVotingAdvocacy Yes, you are right. There are dumb voters who who self weaken their voting power by not putting anybody at the extrema. From data it looks to be around a few percent which is higher than the fraction who give everybody a zero or spoil their ballot. It is more common for voters to put nobody at the MAX than to put nobody at the MIN. One solution is to rescale their ballots as in the first step in Baldwin's method then tabulate as normal using score. This will likely rarely change results but if it does there could be protests. Seems weird to alter ballots anyway so maybe it is just better to tell people what the consequence of doing this is. Another solution is to limit the options with approval voting or to use Baldwin's method itself. There is however, also a reasonable argument to not solve the issue and let dumb voters hurt themselves. --Dr. Edmonds (talk) 20:59, 24 May 2020 (UTC)
- Well, my only suggestion is that cardinal method advocates should not refer to non-normalizing voters as "dumb", even if it is meant to push people to normalize. Probably some good portion of non-normalizing voters would simply be expressing in good faith that their preferences aren't as strong as the other voters', which is more of an altruistic or even "lazy" thing than dumb. I've even heard a game theory type of argument that if you normalize, then you reduce your ability to honestly evaluate the candidates, which in turn makes it harder to incentivize those candidates to better reach out to you. In other words, if you honestly think there is nobody in the field of candidates better than a 2/5 (which could be thought of as an "F" grade), then putting the best of the lot at a 5/5 maximizes your utility gain in that election, but leaves you with no room to increase support for that candidate if they get better later on. I think overall, it may be worth it to have more of a debate or research into what non-normalizing voters are thinking when they score the candidates: maybe there is a dichotomy between those who view it as "giving support for the sake of maximally achieving one's political goals" and those who see it as "grading candidates' quality". BetterVotingAdvocacy (talk) 21:15, 24 May 2020 (UTC)
- BetterVotingAdvocacy. I agree with everything you said there. Dumb is not the right word. "Too honest" is likely better. And it is a good point that if people want to put a protest vote on the ballot giving everybody an F they should be allowed to. I think that research is sorely needed. It could however be that many of the voters did not understand the system. Maybe they did it cumulative or maybe they did not understand that they were weakening their power. The issue is that with this understanding missing it could be used as a criticism of score. I would suspect that whatever the results of such a study show now. If score voting because the standard then the behaviour will change. There is likely a learning curve for score. --Dr. Edmonds (talk) 21:48, 24 May 2020 (UTC)
- Well, my only suggestion is that cardinal method advocates should not refer to non-normalizing voters as "dumb", even if it is meant to push people to normalize. Probably some good portion of non-normalizing voters would simply be expressing in good faith that their preferences aren't as strong as the other voters', which is more of an altruistic or even "lazy" thing than dumb. I've even heard a game theory type of argument that if you normalize, then you reduce your ability to honestly evaluate the candidates, which in turn makes it harder to incentivize those candidates to better reach out to you. In other words, if you honestly think there is nobody in the field of candidates better than a 2/5 (which could be thought of as an "F" grade), then putting the best of the lot at a 5/5 maximizes your utility gain in that election, but leaves you with no room to increase support for that candidate if they get better later on. I think overall, it may be worth it to have more of a debate or research into what non-normalizing voters are thinking when they score the candidates: maybe there is a dichotomy between those who view it as "giving support for the sake of maximally achieving one's political goals" and those who see it as "grading candidates' quality". BetterVotingAdvocacy (talk) 21:15, 24 May 2020 (UTC)
- BetterVotingAdvocacy Yes, you are right. There are dumb voters who who self weaken their voting power by not putting anybody at the extrema. From data it looks to be around a few percent which is higher than the fraction who give everybody a zero or spoil their ballot. It is more common for voters to put nobody at the MAX than to put nobody at the MIN. One solution is to rescale their ballots as in the first step in Baldwin's method then tabulate as normal using score. This will likely rarely change results but if it does there could be protests. Seems weird to alter ballots anyway so maybe it is just better to tell people what the consequence of doing this is. Another solution is to limit the options with approval voting or to use Baldwin's method itself. There is however, also a reasonable argument to not solve the issue and let dumb voters hurt themselves. --Dr. Edmonds (talk) 20:59, 24 May 2020 (UTC)
- BetterVotingAdvocacy It seems the problem is not investigated. I am going to push them to do so in their upcoming paper. I did come across this other paper which as two important results. The first being that when using a scale with negatives voters tend not to give unknown candidates a negative score as they should. This means [0,x] scales should be used. The second result is that different x values produce consistent scoring. This is good to see. --Dr. Edmonds (talk) 18:44, 7 May 2020 (UTC)
- I do not think the consensus candidate argument applies or is even a definable statement in a two candidate race. I think in summary it is possible that voters do vote in lots of crazy ways that ultimately hurt them. Steven Brams says this is why he prefers Approval to Score. he does not trust them to vote properly. It is ultimately an empirical question. Talking about things like grading in the Olympics is not a reasonable analogy. We need examples of actual high stakes political races. I would think the talk tomorrow should provide some empirical evidence. --Dr. Edmonds (talk) 03:17, 5 May 2020 (UTC)
- Wouldn't you consider the common "Score allows a consensus candidate to beat a polarizing majority-preferred candidate" argument to be implying that voters won't necessarily normalize in a two-candidate election, and that indeed, this is a feature of Score, not a flaw? To add, my own argument that voters won't necessarily normalize is that for voters who are wavering between voting and not voting in a particular race, there is nothing "irrational" about them going in the middle and essentially casting a partial vote. BetterVotingAdvocacy (talk) 01:56, 5 May 2020 (UTC)