On voting: How we rank choices, why it fails, and a Jazz Times readers' poll
OK, that's enough of that. I need a break from current events, but I'll combine political science and jazz today by commenting on the basic problem of how we assess collective preferences, why it doesn't work, and a silly poll from Jazz Times.
Yeah, I read Jazz Times. Of course I do. They asked readers to participate in a poll to compile the best jazz albums of the 1990s. Why? 'Cuz. It's a thing that music lovers do. You can find it here, should you care. They are going decade by decade, to do a 50-year spread.
My tastes, of course, are idiosyncratic. To the degree that you care about the list I would make, I spent some time thinking about what it would be, and why, and here's the list I eventually made for my 10 jazz albums for the 1990s. I couldn't actually get it down to 10. I suck. I got it down to 11. Whatever. The list is based on musical vision and originality, which can mean a variety of things, including instrumental approach, structural innovation, and whatever else struck me as I thought about it. I will not claim these as the best, necessarily, because any such list is intrinsically silly, but I will pose this list as a sampling of truly great and innovative jazz albums from the 1990s, any one of which would deserve a place on such a list. Here's where I landed, not in a particular order:
Ginger Baker, Unseen Rain
Henry Butler, For All Seasons
dissent, dissent
Garrison Fewell & Laszlo Gardony, Reflections of a Clear Moon
Bill Frisell, Nashville
Jonas Hellborg, Ars Moriende
Charlie Hunter & Leon Parker, Duo
M-BASE Collective, Anatomy Of A Groove
Sonny Sharrock, Ask The Ages
Steve Tibbetts, The Fall of Us All
Victor Wooten, A Show Of Hands
There's my list, for what it is. For Hellborg, I could have gone with any one of his 1990s albums, but there needed to be at least one, and I limited the list to one. But... my Ginger Baker pick was a collaboration with Hellborg anyway. Should that have sufficed for Hellborg? This is the kind of thing about which I would think, should these things matter, which they don't. Anyway, if you know who these people are, you'll notice a lot of guitarists and bassists because that's my taste, but whatever. It's good music. M-BASE? They deserve recognition for innovation, so let's do that.
Would I really stand behind this list? No, but it's a good list.
So if you examine the options for which Jazz Times lets you vote, most of my choices aren't even options. You get two options for Charlie Hunter albums, but neither are my favorites. You get a Ginger Baker album (Frisell is on that one!), but not my favorite Ginger Baker album. You get two options for Frisell, but not my favorite. I could fill in "other," but what would be the point? A write-in vote will never win. That's... "throwing your vote away!" In fact, the only album on my list that is an option is the Sonny Sharrock option.
Lesson: my tastes are obscure.
So how are people actually voting, as of this morning? Two of the top spots are going to Joe Henderson. That's interesting! So, I seriously dig Joe Henderson. He had a great run of Blue Note albums back in the 1960s, and recorded as a guest with some of the best musicians on some of the most important albums in jazz history. If you aren't heavily into jazz, you may not know him because he wasn't sufficiently famous to get mainstream fame, but in the jazz world, Joe Henderson was big time, and he stuck around for a long time.
If I'm being completely honest, though, he wasn't as vital in the 1990s. Now, it is hard to maintain that level, decade after decade, so this really can't be considered a serious critique. It's just a thing. A lot of that voting is nostalgia voting, and respect for the old giant. It's like giving the Nobel to Einstein for the photoelectric effect because he didn't get it for relativity. Whatever.
Right next to Henderson, in the number 2 spot, you've got a really good album. Pat Metheny's trio album with Dave Holland and Roy Haynes, Question & Answer. Metheny is the weird guy who is, at this point, the old giant still maintaining a level of consistency, but at that point, Holland & Haynes were the old guys bestowing their blessings on him as he hit his stride. I can sort of see that as a pick. Maybe not my pick, but a really good, solid album. Put that on a "Best Jazz Albums of the 1990s," and I'm not going to snark about them giving the Nobel to Einstein for the photoelectric effect.
Among the others, you have obvious critical favorites like Cassandra Wilson's Blue Light 'Til Dawn. You may notice from what I post that I have a bias against jazz singers, and towards instrumentalists, so if I like a singer, that says something. I dig Cassandra, and I think this is her best album. I didn't put it on my list, and I don't think I would, but it is a good album. I will note, though, that Wilson was a member of the M-BASE Collective. Little known fact in jazz history, but it means I had Cassandra on my list anyway. Points in favor of Blue Light: great guest appearances by a range of musicians, including Chris Whitley. As I said, a good album, and not a surprising choice to appear on this list, but it wouldn't have been my pick. If you want to hear modern jazz vocals, though, this is probably where I'll send you.
I could keep going, and comment on the albums with the most votes, but that's not really the point here. The point is, what happens when I look at the structure of choices, knowing something about how that structure, and others' preferences relate to my own? And then, how do we assess the result, because this actually says something about "democracy." Yeah, those are scare quotes.
Let's assume, for the moment, that I care about the results of Jazz Times's poll. If I list Steve Tibbetts, I'm not really taking my primary cues from Jazz Times, but they keep me up to date, and they gave me fodder for this post, so fine, sure, whatever. Anyway, though, the structure of voting does matter more generally, so let's assume I care.
Reference time. Einstein got his Nobel for the wrong thing, but you know who got his Nobel for the right thing? Kenneth Arrow, who won his Nobel Prize in Economics for Social Choice and Individual Values, in which he derived "the impossibility theorem." In colloquial terms, here's what it says. If you have a group of people (e.g. Jazz Times readers) who rank their preferences over a set of options (e.g. jazz albums recorded in the 1990s), it isn't possible to combine their individual rankings into one collective ranking for the group that meets all of Arrow's conditions for democracy and mathematical/logical sense. Sure, you can create a mathematical rule where everyone votes, and you can combine the votes in some way. That will produce an outcome, but the outcome you get will be dependent on the rule you created. A different, and equally defensible voting rule would have produced a different outcome.
Translation: you are being manipulated. It's all about the voting rule. Democracy is an illusion and it can never be otherwise.
Happy Independence Day, everyone!
OK, so let's talk about what's going on with this poll. The rule that Jazz Times is using is as follows: readers check which albums they approve,* and at the end, a top 10 will be created by the number of votes each album receives. As I type, Joe Henderson's Lush Life has 195 votes, with Pat Metheny, Dave Holland & Roy Haynes following at 164. That puts Henderson in the Number 1 slot, and Metheny, Holland & Haynes at Number 2. So that's the rule. And of course, a different rule could produce a different outcome.
How so? Consider this as an alternative: Please check which of these albums you own. Had that been the procedure, I would have checked Blue Light 'Til Dawn, despite my discussion above, and a bunch of others. Isn't this, too, a plausible method? If everyone owns an album, what does that say? You see my point? That's a dramatically different rule, but even subtly different rule changes can have dramatic effects. Why? Because Kenneth Arrow was right.
Given this particular rule, though, suppose I care. What do I do? Remember that the only album on my list that is given as an option is Sonny Sharrock's album, and depending on how many more people vote, it is maybe within reach of the top 10, so I vote for Sharrock, right? OK, fine, done.
After that, it gets tricky. None of my other favorites are even options. So what do I do?
I could write in all of my choices, but they wouldn't do anything. I'm pretty sure I'd be the only one writing in most of that stuff. Maybe someone else writes in Victor Wooten. He has a big name in the bass world, as arguably the greatest living bass guitarist and the only one who could really challenge Jaco for the true title. As Bela Fleck's bass player, even people who don't know bass know him. Steve Tibbetts, though? dissent? Some of that stuff is pretty obscure. I'm doin' nothin' casting those votes.
Wasted votes. Sound familiar? It isn't merely a phenomenon for the plurality rule.
So do I just throw in my lot with Metheny? In the world of constrained choices, that would sound... "sincere," wouldn't it? I could see that. Cassandra? Given the range of options there, that's not a bad choice. Then again, though, it gets messy. We need to address a complex thing in the voting literature. Strategic voting. Voting, not necessarily for your favorite choice, but in order to affect the outcome in the way that you want.
Remember my Charlie Hunter, Bill Frisell and Ginger Baker issues. As I write this morning, there's a Bill Frisell album that is near the top 10, a couple of spots below Cassandra, with Roy Hargrove between. Roy's pretty good too, but I don't own that one. So, maybe I withhold my vote from Wilson and give it to a Frisell album that isn't my favorite Frisell album because Nashville wasn't an option to try to get at least a Frisell album up towards that top 10! You see, the problem is that I'm looking towards the cutoff of the 10th and 11th spots, and if Frisell and Wilson are near each other, I'd prefer Frisell to get it. Have A Little Faith is a good album, even if it isn't my favorite Frisell album, so instead of writing in Nashville, I should vote strategically for Have A Little Faith, and even though I really do like Blue Light 'Til Dawn, I should withhold my vote from it to try to get Frisell up to that 10th spot!
And yes, there are two Charlie Hunter options, but they are so far down in the voting that I wouldn't accomplish much, nor would I accomplish much by voting for Ginger Baker, but that just means I should focus on getting that Frisell album up to the 10th spot, right?
I could elaborate on other aspects of strategic voting here, given the choices, but this demonstrates the basic point. The concept of strategic voting, though, is problematic here, as it has been historically. If the goal of this kind of poll is to measure preferences, and I basically lie about my preferences, then that defeats the purpose, doesn't it? It kinda does. Strategic voting has a long history in voting literature, and it actually predates what we think of as modern democracy because studies of voting, and strategic voting occurred in churches before modern democracy existed. See, for example, George Szpiro's Numbers Rule. And within churches, strategic voting was seen as lying, which was seen as immoral, creating all sorts of problems.
Now, this little survey doesn't really matter, but it does demonstrate the basic mathematical problem. And at the end of the process, we aren't even going to get a "complete" ranking, which means we're in violation of Arrow's conditions anyway. How much should that bother you? If you aren't a technically-minded scholar of voting processes, probably not that much, but here's what you should take away from it, because this does matter. Whenever you observe a voting process, there are a lot of options that get hidden, not by corruption, nor anything like that, but because they have been filtered out through formal political processes. Candidate self-selection, then primaries, and all of that. What you see is a narrow slice, from which you choose. This Jazz Times poll presents a gigantic range of options, with minimal filtering.
Despite that minimal filtering, it has already filtered out nearly all of my favorites, yet still created so much division that even the top spot, which only goes to a sentimental favorite (sorry, Joe), still only gets 2.85% of the vote in a system in which participants cast multiple votes. What will the results tell us? Not a whole lot. Is this because the process is the wrong process? No. It's because there can't be such a process. Not just because all such lists are silly, although they are, but because there cannot be any way to measure collective preferences. There is no such thing as a collective preference.
When you look at a ballot, with choices pre-filtered, the difference between that and the glorious mess with which Jazz Times presented me is that the illusion of order, the illusion of coherence, and the illusion of a mathematical democracy is being presented to you by filtering. Worse, arbitrary filtering.
Pay attention to the man behind the curtain.
*This isn't quite "approval voting," but it is related.
Yeah, I read Jazz Times. Of course I do. They asked readers to participate in a poll to compile the best jazz albums of the 1990s. Why? 'Cuz. It's a thing that music lovers do. You can find it here, should you care. They are going decade by decade, to do a 50-year spread.
My tastes, of course, are idiosyncratic. To the degree that you care about the list I would make, I spent some time thinking about what it would be, and why, and here's the list I eventually made for my 10 jazz albums for the 1990s. I couldn't actually get it down to 10. I suck. I got it down to 11. Whatever. The list is based on musical vision and originality, which can mean a variety of things, including instrumental approach, structural innovation, and whatever else struck me as I thought about it. I will not claim these as the best, necessarily, because any such list is intrinsically silly, but I will pose this list as a sampling of truly great and innovative jazz albums from the 1990s, any one of which would deserve a place on such a list. Here's where I landed, not in a particular order:
Ginger Baker, Unseen Rain
Henry Butler, For All Seasons
dissent, dissent
Garrison Fewell & Laszlo Gardony, Reflections of a Clear Moon
Bill Frisell, Nashville
Jonas Hellborg, Ars Moriende
Charlie Hunter & Leon Parker, Duo
M-BASE Collective, Anatomy Of A Groove
Sonny Sharrock, Ask The Ages
Steve Tibbetts, The Fall of Us All
Victor Wooten, A Show Of Hands
There's my list, for what it is. For Hellborg, I could have gone with any one of his 1990s albums, but there needed to be at least one, and I limited the list to one. But... my Ginger Baker pick was a collaboration with Hellborg anyway. Should that have sufficed for Hellborg? This is the kind of thing about which I would think, should these things matter, which they don't. Anyway, if you know who these people are, you'll notice a lot of guitarists and bassists because that's my taste, but whatever. It's good music. M-BASE? They deserve recognition for innovation, so let's do that.
Would I really stand behind this list? No, but it's a good list.
So if you examine the options for which Jazz Times lets you vote, most of my choices aren't even options. You get two options for Charlie Hunter albums, but neither are my favorites. You get a Ginger Baker album (Frisell is on that one!), but not my favorite Ginger Baker album. You get two options for Frisell, but not my favorite. I could fill in "other," but what would be the point? A write-in vote will never win. That's... "throwing your vote away!" In fact, the only album on my list that is an option is the Sonny Sharrock option.
Lesson: my tastes are obscure.
So how are people actually voting, as of this morning? Two of the top spots are going to Joe Henderson. That's interesting! So, I seriously dig Joe Henderson. He had a great run of Blue Note albums back in the 1960s, and recorded as a guest with some of the best musicians on some of the most important albums in jazz history. If you aren't heavily into jazz, you may not know him because he wasn't sufficiently famous to get mainstream fame, but in the jazz world, Joe Henderson was big time, and he stuck around for a long time.
If I'm being completely honest, though, he wasn't as vital in the 1990s. Now, it is hard to maintain that level, decade after decade, so this really can't be considered a serious critique. It's just a thing. A lot of that voting is nostalgia voting, and respect for the old giant. It's like giving the Nobel to Einstein for the photoelectric effect because he didn't get it for relativity. Whatever.
Right next to Henderson, in the number 2 spot, you've got a really good album. Pat Metheny's trio album with Dave Holland and Roy Haynes, Question & Answer. Metheny is the weird guy who is, at this point, the old giant still maintaining a level of consistency, but at that point, Holland & Haynes were the old guys bestowing their blessings on him as he hit his stride. I can sort of see that as a pick. Maybe not my pick, but a really good, solid album. Put that on a "Best Jazz Albums of the 1990s," and I'm not going to snark about them giving the Nobel to Einstein for the photoelectric effect.
Among the others, you have obvious critical favorites like Cassandra Wilson's Blue Light 'Til Dawn. You may notice from what I post that I have a bias against jazz singers, and towards instrumentalists, so if I like a singer, that says something. I dig Cassandra, and I think this is her best album. I didn't put it on my list, and I don't think I would, but it is a good album. I will note, though, that Wilson was a member of the M-BASE Collective. Little known fact in jazz history, but it means I had Cassandra on my list anyway. Points in favor of Blue Light: great guest appearances by a range of musicians, including Chris Whitley. As I said, a good album, and not a surprising choice to appear on this list, but it wouldn't have been my pick. If you want to hear modern jazz vocals, though, this is probably where I'll send you.
I could keep going, and comment on the albums with the most votes, but that's not really the point here. The point is, what happens when I look at the structure of choices, knowing something about how that structure, and others' preferences relate to my own? And then, how do we assess the result, because this actually says something about "democracy." Yeah, those are scare quotes.
Let's assume, for the moment, that I care about the results of Jazz Times's poll. If I list Steve Tibbetts, I'm not really taking my primary cues from Jazz Times, but they keep me up to date, and they gave me fodder for this post, so fine, sure, whatever. Anyway, though, the structure of voting does matter more generally, so let's assume I care.
Reference time. Einstein got his Nobel for the wrong thing, but you know who got his Nobel for the right thing? Kenneth Arrow, who won his Nobel Prize in Economics for Social Choice and Individual Values, in which he derived "the impossibility theorem." In colloquial terms, here's what it says. If you have a group of people (e.g. Jazz Times readers) who rank their preferences over a set of options (e.g. jazz albums recorded in the 1990s), it isn't possible to combine their individual rankings into one collective ranking for the group that meets all of Arrow's conditions for democracy and mathematical/logical sense. Sure, you can create a mathematical rule where everyone votes, and you can combine the votes in some way. That will produce an outcome, but the outcome you get will be dependent on the rule you created. A different, and equally defensible voting rule would have produced a different outcome.
Translation: you are being manipulated. It's all about the voting rule. Democracy is an illusion and it can never be otherwise.
Happy Independence Day, everyone!
OK, so let's talk about what's going on with this poll. The rule that Jazz Times is using is as follows: readers check which albums they approve,* and at the end, a top 10 will be created by the number of votes each album receives. As I type, Joe Henderson's Lush Life has 195 votes, with Pat Metheny, Dave Holland & Roy Haynes following at 164. That puts Henderson in the Number 1 slot, and Metheny, Holland & Haynes at Number 2. So that's the rule. And of course, a different rule could produce a different outcome.
How so? Consider this as an alternative: Please check which of these albums you own. Had that been the procedure, I would have checked Blue Light 'Til Dawn, despite my discussion above, and a bunch of others. Isn't this, too, a plausible method? If everyone owns an album, what does that say? You see my point? That's a dramatically different rule, but even subtly different rule changes can have dramatic effects. Why? Because Kenneth Arrow was right.
Given this particular rule, though, suppose I care. What do I do? Remember that the only album on my list that is given as an option is Sonny Sharrock's album, and depending on how many more people vote, it is maybe within reach of the top 10, so I vote for Sharrock, right? OK, fine, done.
After that, it gets tricky. None of my other favorites are even options. So what do I do?
I could write in all of my choices, but they wouldn't do anything. I'm pretty sure I'd be the only one writing in most of that stuff. Maybe someone else writes in Victor Wooten. He has a big name in the bass world, as arguably the greatest living bass guitarist and the only one who could really challenge Jaco for the true title. As Bela Fleck's bass player, even people who don't know bass know him. Steve Tibbetts, though? dissent? Some of that stuff is pretty obscure. I'm doin' nothin' casting those votes.
Wasted votes. Sound familiar? It isn't merely a phenomenon for the plurality rule.
So do I just throw in my lot with Metheny? In the world of constrained choices, that would sound... "sincere," wouldn't it? I could see that. Cassandra? Given the range of options there, that's not a bad choice. Then again, though, it gets messy. We need to address a complex thing in the voting literature. Strategic voting. Voting, not necessarily for your favorite choice, but in order to affect the outcome in the way that you want.
Remember my Charlie Hunter, Bill Frisell and Ginger Baker issues. As I write this morning, there's a Bill Frisell album that is near the top 10, a couple of spots below Cassandra, with Roy Hargrove between. Roy's pretty good too, but I don't own that one. So, maybe I withhold my vote from Wilson and give it to a Frisell album that isn't my favorite Frisell album because Nashville wasn't an option to try to get at least a Frisell album up towards that top 10! You see, the problem is that I'm looking towards the cutoff of the 10th and 11th spots, and if Frisell and Wilson are near each other, I'd prefer Frisell to get it. Have A Little Faith is a good album, even if it isn't my favorite Frisell album, so instead of writing in Nashville, I should vote strategically for Have A Little Faith, and even though I really do like Blue Light 'Til Dawn, I should withhold my vote from it to try to get Frisell up to that 10th spot!
And yes, there are two Charlie Hunter options, but they are so far down in the voting that I wouldn't accomplish much, nor would I accomplish much by voting for Ginger Baker, but that just means I should focus on getting that Frisell album up to the 10th spot, right?
I could elaborate on other aspects of strategic voting here, given the choices, but this demonstrates the basic point. The concept of strategic voting, though, is problematic here, as it has been historically. If the goal of this kind of poll is to measure preferences, and I basically lie about my preferences, then that defeats the purpose, doesn't it? It kinda does. Strategic voting has a long history in voting literature, and it actually predates what we think of as modern democracy because studies of voting, and strategic voting occurred in churches before modern democracy existed. See, for example, George Szpiro's Numbers Rule. And within churches, strategic voting was seen as lying, which was seen as immoral, creating all sorts of problems.
Now, this little survey doesn't really matter, but it does demonstrate the basic mathematical problem. And at the end of the process, we aren't even going to get a "complete" ranking, which means we're in violation of Arrow's conditions anyway. How much should that bother you? If you aren't a technically-minded scholar of voting processes, probably not that much, but here's what you should take away from it, because this does matter. Whenever you observe a voting process, there are a lot of options that get hidden, not by corruption, nor anything like that, but because they have been filtered out through formal political processes. Candidate self-selection, then primaries, and all of that. What you see is a narrow slice, from which you choose. This Jazz Times poll presents a gigantic range of options, with minimal filtering.
Despite that minimal filtering, it has already filtered out nearly all of my favorites, yet still created so much division that even the top spot, which only goes to a sentimental favorite (sorry, Joe), still only gets 2.85% of the vote in a system in which participants cast multiple votes. What will the results tell us? Not a whole lot. Is this because the process is the wrong process? No. It's because there can't be such a process. Not just because all such lists are silly, although they are, but because there cannot be any way to measure collective preferences. There is no such thing as a collective preference.
When you look at a ballot, with choices pre-filtered, the difference between that and the glorious mess with which Jazz Times presented me is that the illusion of order, the illusion of coherence, and the illusion of a mathematical democracy is being presented to you by filtering. Worse, arbitrary filtering.
Pay attention to the man behind the curtain.
*This isn't quite "approval voting," but it is related.
Comments
Post a Comment