ENnies discussion thread

[Dextra]

Voting Mechanics

First of all, I'd like to thank those of you who are trying to educate us neo-Luddites in the mechanisms and vagaries of polling systems. I'm trying to wade through the information as best I can, and I'm sure between the ENnies Board of Directors and Michael Morris, with the help of such fine minds of ENWorlders, we'll have a truly excellent system.

Jester47- technical details should be discussed with Michael Morris, although you're welcome to CC me in the conversations so I can pretend I understand what you're talking about.

I have a question for you voting expert peeps.
Last year we rated products on a scale, giving the chance to opt out if you weren't familiar with a given product. This is generating some discussion and apparently some consternation.
What about the idea of rating products relative to each other- ie., there's five products, mark your fave as number one, and least fave as number five, and so on, then assign points to each value, and give the award to the top two scorers? Please forgive my ignorance, and try to be gentle as you explain this to me.

 

[Conaill]

What about the idea of rating products relative to each other- ie., there's five products, mark your fave as number one, and least fave as number five, and so on, then assign points to each value, and give the award to the top two scorers? Please forgive my ignorance, and try to be gentle as you explain this to me.

If you have to give an exact ranking, that is Borda Count - see Fusangite's discussion above why it's not a great system.

There are other systems that also use a ranking and don't lead to the same type of problems - such as Instant Runoff Voting. the input from the voter is the same: a ranking from 1-5. They only differ in how you use those votes to figure out who wins.

In principle, a ranked voting method should allow us to deal with "don't know" votes more easily: if there is one product you are not familiar with, just rank the remaining 4 and leave the unknown one unranked. In practice, such unranked candidates can make the system a lot more complicated.

 

[Umbran]

Approval Voting shares many of the beneficial properties of Alternative Vote (IRV, Preferential Voting, single-winner STV, whatever ) and has some additional nice features, but is much easier to implement and therefore often more readily accepted by voters. In our particular case, it also has an easier time dealing with the many "don't know" votes.

AV is simpler to implement, yes. But it isn't so robust in dealing with ties.

IRV has the benefit that it measures not only strength of preference, but bredth as well. Being the top preference of many is important, but if many people also like it as second and third, that may also show.

IRV doesn't have a problem with "don't know" votes, or with "I disliked this so much I don't want to vote for it at all". YOu simpl;y don't give a ranking to products you don't know. If you only know one product, and you think it deserves the prize, you rank it first and don't rank the rest. Your ballot won't count after your preference is eliminated, but then, you are admitting you don't know enough to have a second preference anyway.

 

[Conaill]

AV is simpler to implement, yes. But it isn't so robust in dealing with ties.

IRV has the benefit that it measures not only strength of preference, but bredth as well. Being the top preference of many is important, but if many people also like it as second and third, that may also show.

I'm not sure what you mean with "less robust in dealing with ties". Care to elaborate?

As for breadth of support, it is a well-known issue that IRV may eliminate a centrist candiate too ealy, if he gets insufficient 1st-choice votes. If anything, Approval Voting is more geared towards a consensus candidate.

IRV doesn't have a problem with "don't know" votes, or with "I disliked this so much I don't want to vote for it at all". YOu simpl;y don't give a ranking to products you don't know. If you only know one product, and you think it deserves the prize, you rank it first and don't rank the rest. Your ballot won't count after your preference is eliminated, but then, you are admitting you don't know enough to have a second preference anyway.

Note that with this scheme, not ranking a product is equivalent to ranking it *last*. (Correction: it's actually even worse than that if you have multiple unranked products.) That seems hardly fair to the products you don't know! (And is precisely the situation they were trying to avoid last year.) It is possible to deal with unkowns in a more fair way - essentially by extrapolating from the rankings other voters gave to the product - but that complicates the system quite a bit.

In Approval Voting, you are asked to draw a line between one set of candidates and the rest. This implicitly forces you to make an estimate of the quality of the unknown candidates. For example, the ballot might consist of two ok products, one really crappy one and two you don't know. Under Approval Voting, you may decide to vote for the two ok products and the two you don't know, under the assumption that any unknown Ennies nominee will likely be better than the one product you know is crap. This is an entirely reasonable and acceptable vote, based on your estimate of the quality of the candidates. Or if one of the two "ok" products is instead a superb work of art, you may decide to vote only for that one, under the assumption that the two unknown are likely not quite as brilliant. Again, a perfectly valid vote.

The upshot is that Approval Voting will give an unbiased estimate of the quality of partially unknown candiates, without having to add any mechanism to deal with them explicitly.

 

[Macbeth]

I'd say the best way would be to treat folks as if they were mature and thoughtful - talk about the various merits of having long-term judges and new-blood judges, and let the results come out in voting.

I too think this is best. People trust people like Crothian and TB (and for good reason). I don't think the voting system should force anybody to vote for a second choice, just to keep things fresh.

On the other hand, I have great confidence inthe EN World populace to be willing to give new jdges a chance, especially if we give them reason to. I want to show the populace that I could be a judge by posting reviews, so I made sure I got my Hamunapatra review up this week. If other canidates want to see change in the judges, I suggest the same thing: make yourself obvious, and show that you can do the job. I'd love to give my point of view to as a judge, so I'm trying to make it clear that I have something to give.

 

[Umbran]

I'm not sure what you mean with "less robust in dealing with ties". Care to elaborate?

As you describe it, Approval Voting can yield ties. Not just between two leaders, but between any number of candidates. The ballots contain little information you can use to resolve the tie, and you must make assumptiosn about the voter's intent in order to use it.

Basic Instant Runoff Voting can also have ties, but the ballots contain further information you can use to resolve them without making assumptions about the user's intent.

As for breadth of support, it is a well-known issue that IRV may eliminate a centrist candiate too ealy, if he gets insufficient 1st-choice votes. If anything, Approval Voting is more geared towards a consensus candidate.

Oh? If any one candidate gets more than 50% of the vote on the first pass, either IRV or AV will show it. If a candidate gets less than that consensus, AV will still give the win to the leader. With AV and five candidates, a winner can have just over 20% of the people voting for it. You want to claim that 21% is a consensus?

It is precisely this problem that IRV is designed to avoid. If the consensus is not clear on the first pass, it has other information on hand about how much people like the products.

In Approval Voting, you are asked to draw a line between one set of candidates and the rest. This implicitly forces you to make an estimate of the quality of the unknown candidates.

If it is unfair to rank an unknown last, it is even more unfair to rank an unknown ahead of a known. In essence, you're saying that people get to (are even forced to) lie, and say they like an unknown better or worse.

This is an entirely reasonable and acceptable vote, based on your estimate of the quality of the candidates.

If I was in a shop, planning to buy a product with my own money, it might be reasonable to estimate the quality of an unknown. But it is not reasonable for an awards program.

The upshot is that Approval Voting will give an unbiased estimate of the quality of partially unknown candiates, without having to add any mechanism to deal with them explicitly.

"Unbiased" does not mean "accurate" or even "fair". I think that the estimate here actually can contain a bias. Any time you estimate a quality, you can introduce bias, based upon how you make your estimate.

Say there's three products - the voter liked A, disliked B, and does not know C. If he thinks to himself, "A number of other things from C's publisher were good, so it must be better than B", and approves A and C, the voter has just introduced a bias for a product that has nothing to do with the product itself. That isn't fair.

This is a problem with approval voting - it has only two states, and that forces people to vote as if they had knowledge they do not.

 

[Dextra]

More voting issues

Okay stats/votes guys, here's some more questions for you-

The two methods you've been discussing seem to be geared for choosing a single winner.
How do they compare for picking a first and second place?

Michael- can either be implemented using our current voting system? Could you program one of those methods? Or would we have to tally votes by hand?

We'll be including a Fan's Choice award which will be a write-in option, so there will have to be some sort of human tallying anyhow.

Also, I was wondering if we should consider making a request that if you are not familiar with at least two (or maybe three?) of the products that you should abstain from voting in the category altogether? You can't really say that a certain product is the best if it's the only one you've ever seen!

 

[Conaill]

As you describe it, Approval Voting can yield ties. Not just between two leaders, but between any number of candidates. The ballots contain little information you can use to resolve the tie, and you must make assumptiosn about the voter's intent in order to use it.

You mean ties where two or more candidates get exactly the same number of votes? That's a non-issue in practice. Such ties can happen with any voting method, but are exponentially unlikely with a reasonable number of voters. (And I'm sure Morrus wouldn't mind handing out two Gold Ennies in that case. )

With AV and five candidates, a winner can have just over 20% of the people voting for it. You want to claim that 21% is a consensus?

Keep in mind that Approval Voting encourages people to vote for multiple candidates, so the situation you just described will never happen. (The exact same situation can happen in IRV if everyone only ranks one choice per ballot - but that'll never happen in real life either.)

Maybe I should have used "centrist" instead of "consensus", i.e. a candidate which most of the voters approve off, although they may not pick him as their first choice. Approval Voting tends to encourage such "middle of the road" candidates. (This is actually one of the more substantial arguments against it for political elections, because it discourages politicians from taking a stand.) IRV on the other hand encourages strong contrasts, because most of the power is located in the first-choice votes (the vast majority of lesser choices in the votes never actually comes into play). IRV may eliminate a centrist candidate early on in favor of a polarized choice that enjoys smaller actual support. Again, these are very well known and often discussed properties of Approval Voting and IRV.

Some other known disadvantages of IRV:

(1) nonmonotonicity: under some circumstances, ranking a candidate higher may actually cause him to lose

(2) the 3rd-party spoiler effect: although IRV is often touted as a solution to the spoiler effect and a boon for 3rd parties, this only holds as long as said 3rd party is unlikely to win. As soon as you get more than two plausible winners, the spoiler effect returns, causing one to vote for the "lesser of two evils". (This is probably why Australia is still stuck with a 2-party system, even though it has used IRV since the 20's or so.)

Some more endorsements of Approval Voting;

- It is used by the American Statistical Association, the Mathematical Association of America, the Institute of Electrical and Electronics Engineers (>300K members), the US National Academy of Sciences, the United Nations, as well as numerous smaller professional organisations and universities.

- "The Science of Elections" S.J. Brams and D.R. Herschbach, Science May 25, 2001: 1449.

- Scientific American "Ask the Experts" discussion (two leading mathematicians in favor of Approval, one for (gasp) Borda Count.

- Some FAQ questions on IRV vs Approval Voting

If it is unfair to rank an unknown last, it is even more unfair to rank an unknown ahead of a known.

Why?

Seriously - why would it be unfair to assume that a product you are familiar with and have a strong dislike for is actually *worse* than another Ennie nominee you are not familiar with? After all, these nominations have been through an exstensive vetting process already, so you know *something* about their quality. Why would it be more fair to assume that the unknown product is even worse than the one you know you detest? It's not as if you're forced to stick an exact value on the unknowns.

In statistics, that sort of estimated quality of an unknown entry is called a prior, and it's perfectly acceptable to have a prior estimate which is higher than one of your known datapoints.

 

[Rasyr]

I think I will sneak in a comment...

The way I see it (with my limited understanding)
The Approval Voting system woulld/could pick multiple winner through a straight percentage of how many votes a products got from the total number of votes cast. However, if there is a tie, or you are wanting a minimum percentage of votes for the front-runner to be declared a winner, this system does not support that very well.

With the Instant Run-off Voting, it appears that this would work much much better, and can deal with requiring a minimum percentage for a win (within certain limits). If I am understanding it properly. It would work as follows....

Take the following example of 6 books with 600 people voting. Each voter ranks the products from 1 - 5 (with a possible null vote as well). And say we end up with a table that looks like the following:

[b]Product	Total votes for first			[/b]
Book 1	175	
Book 2	150	
Book 3	100	
Book 4	75	
Book 5	60	
Book 6	40

In the above example, there is no winner (because none got over 50%). So Book 6 would be dropped, and the 40 votes that went to Book 6 would be redistributed among those according to their second choice. (these second choices are now considered first choices, and the third choice is now considered second) This would then give us the followling:

[b]Product	Total votes for first	[/b]
Book 1	180	
Book 2	175	
Book 3	100	
Book 4	85	
Book 5	60

Still no winners (over 50%), So we now drop Book 5 and redistribute those 60 votes according to what they had ranked as their next choice.

[b]Product	Total votes for first	[/b]
Book 1	190	
Book 2	200	
Book 3	110	
Book 4	100

Still no majority, so we rinse and repeat.

[b]Product	Total votes for first	[/b]
Book 1	210	
Book 2	240	
Book 3	150

Still nobody with a 50% majority, so we now drop Book 3 and try again after redistributing those 150 votes accoriding to their next choice.

[b]Product	Total votes for first	[/b]
Book 1	295	
Book 2	305

And book 2 is the winner.. Now, IF there were a tie, then my suggestion would be to go back to the original vote and the one with the largest number of original 1st votes would be the winner (i.e. Book 1 IF it had been a tie).

I pretty much ignored the nulls that may have been in there (from folks who voted a null because they did not know the product).

I hope that helps Dextra. As to implementing it.... Ask Mr. Morrus.

 

[fusangite]

Glad to see AV, now called IRV in this thread to avoid confusion with approval voting, was explained in my absence.

In my view, the main thing we should look for in a system is this: if people engage in true hardball strategic voting to make their favourite product win, will it result in people giving positive votes to products they think are bad and negative votes to products they think are good? If yes, don't use the system. If no, we can move forward. While it will take me some time to digest the particular brand of approval voting the non-profit society to which Conaill refers us favours, my initial inclination is to question whether this particular brand of approval voting meets these criteria.

Many credible political scientists and mathematicians who work on voting systems these days oppose approval voting systems precisely because, like Borda-style systems, they are unable to cope with large-scale strategic voting. Approval voting, when used in electoral processes, tends to lead to highly negative campaigns because engendering disapproval of opposing candidates is behaviour that the system rewards. In the first duma election under Gorbachev, approval voting resulted in a majority of the seats being unoccupied at the end of the election because the various negative campaigns were so successful that in most districts, none of the candidates managed to avoid majority disapproval.