Time to vote on this week's blog posts! · 2:12am Nov 17th, 2015
It's time for all of you who chose the "vote on what Bad Horse should blog about this week" Patreon reward to do so!
And that is... Blagdaross. Can I still call it "voting" when there's only one voter?
EDIT: Scratch that. I'm going to extend Bad Suffrage to everypony who got the $5 or $8 reward as well. That's equestrian.sen, horizon, Forderz, Bradel, and Blagdaross.
Remember, "vote" means "comment on the post and list the ones you want me to work on," and I work on the one with the most votes. Optimal behavior is probably to choose half of the topics. Maybe equestrian.sen can figure it out for sure.
The blog topics I've been thinking about recently are nearly all related to my study of the history of literature and literary theory. Some of these may need several other introductory blogs first. Choosing what order to write about these things is hard. There's no single starting point.
Well, maybe there is one. Homer's Iliad. So a likely early blog post is "The Rage of Bad Horse: Review of Homer's Iliad." I did not like the Iliad, though this isn't entirely fair, since it was designed as oral poetry, not as literature. It gets high marks, though, for integrating action, drama, and philosophy in the same vision, which literature doesn't do much anymore.
"How dead white guys control your TV and radio": A mini-study of how philosophy affects literature. Not at all comprehensive. There are probably lots of people who know a lot more about this; I just haven't found them. Probably I should take a few more weeks to work on this one.
"Why it's Easy to Write for MLP": A comparison of the plots of different TV shows, and one distinctive feature of MLP plotlines.
"Tics, Stereotypes, and Character": A short, inconclusive piece musing whether a "round" or "deep" character is anything more than a sufficiently large pile of cheap tricks.
"The Quixote gutted by literary fashion": Rage at Grant Voth's discussion of the perspectivist interpretation of Don Quixote.
"The New Sincerity": Making stuff un-ironically good is now considered a radical new literary movement.
"Pitirim Sorokin's Social & Cultural Dynamics": Pitirim Sorokin published a MASSIVE book in 1937--it was something like 2000 pages--whose basic thesis was that there are two types of cultures, good idealistic ones, and degraded sinful ones. He used different words, but that was what he meant. The first quarter of the book was a review of art history, and I mean all the art history he could review. Mainly Europe from Homer to 1930, but also caveman art, Asia, Egypt, and a few other times and places where he could. His main conclusion was very similar to mine, which is why I got the book. He says that there's one principal dimension of art (though he didn't use that term, since only sinful empiricists count things) between idealism and naturalism/empiricism, and over and over again, we find a whole host of specific properties of art are determined by what position a culture is at along that dimension.
His analysis is insightful in some ways, blind in others. What's most illuminating about it is that the slow piling up of craziness helped me understand, I think, that people really liked the things they said they liked in the Middle Ages, and showed me how their beliefs warped not just their thoughts, but their perceptions and feelings, just like Sorokin said beliefs do. It's very hard to wrap my head around without dismissing them as just stupid and crazy. They were stupid and crazy, and so was Sorokin, but it was an internally-consistent stupid and crazy.
Unfortunately, his data is in the 1937 edition, which is hard to get and expensive when it's available. I have the 1957 edition, which is now $11.95 + shipping on Amazon. Get it now if you want to read it before I blog about it.
"Naturalist versus Idealist Poetry": A comparison of the anonymous medieval Scottish poem "Edward" with John Keats' romantic, idealistic poem "La Belle Dame Sans Merci". The later poem is an example of idealistic poetry; the older poem, of naturalistic. See why I think idealistic poetry sucks.
"J. S. Bach and the discovery of ugliness": A closely-related post on idealistic versus naturalistic art. Bach wasn't considered great during his lifetime, but is now. I'll suggest this is because Bach had a modern sense of the sublime, as beauty plus ugliness, in an age when unpleasantness and disquietude were still banned from art, and art was mostly weak sauce as a result.
"The Creativity Revolution": The idea that humans could be creative didn't exist in the 17th century. Evidence for this, an explanation, why it changed after 1700, and the results. This may take more than a week to finish.
"Review of Martindale's The Clockwork Muse": Colin Martindale published The Clockwork Muse: The Predictability of Artistic Change in 1990. Right now it's 40 cents plus shipping on Amazon, so now would be a good time to buy it if you want to read it before I publish my blog on it. Martindale also thought he'd found the principal component of art, which was "primordial content" as a function of time. He believed that artistic change is internally driven by the quest for novelty. So an art moves regularly through different styles, and within each style, the fraction of primordial content goes down, then up.
He did loads of quantitative, statistical tests in many different art forms, covering the past 700 years, to prove this. Unfortunately, he didn't understand statistics, and his analysis is plagued by underpowered analysis, a failure to account for degrees of freedom, and post-hoc rationalization. But it shows (mostly) how to do quantitative art experiments, and would be worth following up in a more mathematically correct manner. Some of his conclusions are still valid, such as his proof that the terms "baroque", "romantic", and "classical" mean something other than just "what people did during a specific time period", something that holds across music, painting, sculpture, and architecture. This may take more than a week to finish.
"The Principal Dimension of Art": I'm leading up to this. And I still want the data from Sorokin's 1937 edition. I'm going to put this on hold until 2016.
"Mushrooms" is the name of a poem by Sylvia Plath. It's a good poem, but what's more interesting to me is that we have instances of professional poetry critics who completely failed to get what it was about. We also have data from an English class showing that most students also completely missed the point. And it isn't even an obscure poem. I'll probably merge this with a blog on Eudora Welty's comments on one of her short stories, and some comments by Fred Millett in Reading Poetry, showing the vast gap between what readers understand and what most writers, editors, and teachers assume they'll understand. Also some wordage on the survival of the author: Authors really do have an intent, and people who find something else, usually find something stupid.
"Writing in e-prime": Why it's a good idea to try it. Probably not the reason you think. E-prime means English without the verb "to be".
"Writing: Good ambiguity, bad ambiguity": My thoughts on why or when ambiguity can be good. This would take a long time to finish. I started it before Bronycon.
"The Greatness of Norman Rockwell": Why Norman Rockwell is a great artist. This one probably has to go after "Naturalist versus Idealist Poetry", "Review of Martindale's The Clockwork Muse", and maybe the review of Pitirim Sorokin.
"An appraisal-based ontology of fiction": A way of categorizing stories based on sentiment analysis.
"Dramatica": Dramatica is a really interesting and slightly nutso theory of how stories work. If you choose this, all I can do this week is spend more time studying it. I've already put a dozen hours into this, and it will take dozens more. You can download a PDF of the free online Dramatica book here, or read it online here.
"The Mystery of Mysteries": I think that each genre has a debased form, which is what we usually think of the genre as being, and at least one original meaningful form. My post "Fantasy as deontology" was about the meaningful form of fantasy. I started thinking about this post because I kept reading people saying that, of course you don't need to worry about character if you're just trying to write a Sherlock Holmes story. This is a stupid thing to say, and suggests that Sherlock Holmes stories aren't about what most people think they are. At first I thought this was peculiar to Sherlock Holmes, but now I think it's a more general story pattern that good mysteries usually follow. This will take at least three full days to write, because I need to read more mysteries.
Choose wisely, ponies! And be careful what you wish for. 
[This post doesn't count towards a 10K-word unit.]
Yes, choose wisely, Blagdaross and equestrian.sen! I'll be here holding a wine glass with my pinky finger stuck out, and snickering behind my hand that I spent too much to have a vote like you proletariat.
All democracies are based on the simple principle of "One man, one vote." I'm the man. I get the vote.
This one seems extra interesting to me - both in terms of writing literature and in terms of playing games. Throw in some film and I'll be tickled pink.
3546871 You might also want to get the $5 "View my story drafts" reward. It would come in helpful for your $8 "Comment on my story drafts" reward.
Seriously, though, that's good use of strategic guilt. Is your mother from the old country, maybe?
3546912 Sorry, I lied. I'm gonna make the rewards for repeat pledges cumulative. That doesn't include the one-time "choose my avatar" pledge, but the $5 and $8 pledges will also come with blog voting. That makes it you, eq-sen, horizon, and Forderz.
3546933 Very well. I shall somehow find a way to tolerate their ... opinions.
My vote stands, though. That's right up my alley.
3546947 Okay, but as I tried to warn you, so far I haven't come up with much to say about it.
I've got to admit, my first thought on Mushrooms was that it was about bland, boring, uninteresting people becoming more and more prevalent and overrunning society with their blandness, not through any action of their own, but simple inertia.
Not sure if that's right or not. Looking at some analysis online, I see people arguing everything from it being about women gaining more of a role in society (which seems potentially reasonable; mentions needles, with those being traditionally associated with women, and the whole bland-mannered, shoving and pushing in spite of themselves would be traditionally unlike women; on the other hand, it doesn't exactly seem like a rousing women's lib poem, and it wasn't like women's lib was exactly quiet) to it being about the Cold War (because mushrooms have to be about nukes, right? Mushroom clouds? I'm pretty confident that interpretation is outright wrong, given that I would never call nukes "bland-mannered").
To mangle Pratchett here:
“Equestria had dallied with many forms of government and had ended up with that form of rule known as One Pony, One Vote. The Princess was the Pony; she had the Vote.”
If your blog topics were porn I'd masturbate to them, if that makes any sense.
images2.wikia.nocookie.net/__cb20121201234624/mlp/es/images/f/fa/Trixie_innocent.png
3547376 Um... just try not to touch anything, okay?
3546912
3547613
ultimateclassicrock.com/files/2013/01/96604940.jpg
3546933
Strategic guilt is the tactical nuke of interpersonal relations. I've been working for quite some time to steal it and hold the world hostage. I learn, after all, from the baddest.
Speaking of which, I think the evilest vote I could cast is for all of your topics above equally. This has the inconvenient side effect of preserving 3546912's extra voting power this week, or perhaps, depending on how I frame it, the convenient side effect of allowing me an additional week of extended-pinkie-finger snobbery.
I do, however, demand a public accounting of what happened to the unkicked puppy.
vignette3.wikia.nocookie.net/undertale/images/8/8b/ExpandDog.png/revision/latest?cb=20151030064123
3548197
Oh, the puppy! Thank you for reminding me. I have a post to post on that.
Oh God. The Emo kids have graduated.
3548872 I think it refers to the idea that you don't have to be ashamed of trying to do something good. See the wikipedia entry.
I think that depends on whether or not we're selecting from a countable number of blogs.
Wait, that doesn't sound right... The only winning move is not to play. Yeah, let's go with that.
We can convey enough, and varied enough, information in other ways (e.g. a comment) that the information conveyed through our votes can be made irrelevant. It would be suboptimal to use the votes to convey any amount of almost any kind of information for information that the voter actually wants to convey. Of course, people are going to attribute a disproportionate amount of meaning to the votes regardless of how inefficient that might be as a voter, so forget that.
I think the only "optimality pressure" changing a voter's strategy is a voter's impact on other voters. Swaying people should prefer voting for few blogs, while anti-swaying people should prefer voting for a lot of blogs. The aim is to maximize the chance that others will vote along with your preference, and minimize the chance that others will vote against your preference.
Ignoring the case where a voter actually prefers exactly half the blogs, I don't think there's any "optimality pressure" that pushes voters towards that particular fraction (50%). If you don't know what effect you'll have, then I think the optimal strategy is to not account for voter sway and just vote naturally, whatever that means.
"The Creativity Revolution"
"The Mystery of Mysteries"
3550106 Let me be more specific: Suppose your utility function for each topic has a normal distribution. You don't actually know the utility for each; you can only rank all the topics from highest utility to lowest. You can only vote yes or no for each topic. ("No" and not voting are equivalent.) All voters have uncorrelated utilities. How many should you vote for to maximize your expected utility?
3550440 I can't say I've ever seen a discussion on blog post topics devolve into some twisted form of game theory.
3550627 3550106 All of our conversations are game theory. We just don't usually admit it.
3550926
3550440
That doesn't actually change the problem. Assuming no correlation of utility across topics, you should always vote "yes" where the mean utility is greater than zero, and "no" where the mean utility is less than zero. You can vote arbitrarily where the mean utility is zero.
Suppose voting "yes" for a topic sets the probability that the topic will be chosen to α. Voting "no" for a topic sets the probability that it will be chosen to β. The expected utility by voting "yes" for any given topic is then the infinite integral of U*p*α over the utility distribution function, where p is probability density of the utility distribution function at U. Since α is a constant, this simplifies to the expected utility of the topic multiplied by α, E*α. Similarly, the expected utility by voting "no" for any given topic is E*β.
Since utility across topics is uncorrelated, utility is maximized for the set of decisions when utility is maximized for each decision individually. If you want to maximize expected utility for any individual decision, you want to vote "yes" if E*α is greater than E*β, and "no" if the reverse is true. If we assume that voting for a topic does not make it less likely to be selected, then α >= β, so you want to select "yes" when E is greater than 0 and "no" when E is less than 0. When E is 0, the decision is arbitrary since there will be no difference in expected utility in either case.
Of course, this all assumes that personal preference is not an indicator of global preference. If it is, then α and β are both dependent on the utility distribution functions, and the simplification of the integrals to E*α, E*β expected utilities is incorrect.
Edit:
Aaaaaaaand I missed that. You have to also assume that nothing about the mean of all topics is known, or that the mean is known to be zero. You also have to change your assumption from "all voters have uncorrelated utilities" to "all voters have uncorrelated utility distributions" since uncorrelated utilities (assuming the traditional Pearson correlation) can still have correlated means. In this case you want to select "yes" for the ones on the high end and "no" for the ones on the low end, which would mean selecting 50% of the topics.
3555016
I don't understand what U represents in your equation. If you mean to integrate with respect to U, you can't; you're voting on points, not on a continuum.
But in any case, if correct, this would apply even when you're the only voter, in which case it's obviously false. If you're the only voter, you should make just one vote, for your highest-utility topic. You only want to vote for everything with above-zero utility to you if there are so many votes that your vote changes the probability of each one winning infinitesimally.
If there are 2 voters, it isn't so obvious what to do.
Suppose there are 2 voters. You vote for n-1 of the N topics, starting with the one with highest utility. Do you vote on topic n?
Voting on topic n increases utility if it causes topic n to have 2 votes (other voted for n) and the other voter didn't vote for any topic in n+1 ... N. It decreases utility if the other voter did vote for any topic in n+1 ... N. We don't have to know the utilities; just the rank order.
To be solvable, the problem may need to assume that all voters are perfectly rational. Hence, you and the other voter will both vote for v topics. Then we have (v/N) * [(N-v)/N]^v (probability topic v got a vote from the other * probability no topics > v did) = 1 - [(N-v)/N]^v, ((v+N)/N) * [(N-v)/N]^v = 1. Wolfam Alpha, solve (v/N + 1) = [N / (N-v)]^v :
www5a.wolframalpha.com/Calculate/MSP/MSP36441fi96c2a5g0dd7bc00004b37ef139209agcf?MSPStoreType=image/gif&s=24&w=300&h=224&cdf=RangeControl
I don't know what the solution is, but it looks like N has to be pretty large before v goes up to 2. This analysis is flawed because it doesn't account for what happens when there's a tie. My intuition is v should be set so that ((N-v)/N)^v ~ .5, so you have about an even chance of one of the things you vote for being chosen. Note that in this case N=1 => v ~ 1.2, which is pretty close to 1, as desired.
www5a.wolframalpha.com/Calculate/MSP/MSP2961fd2fch03d23b61h00003735f63di629f0g6?MSPStoreType=image/gif&s=33&w=200&h=150&cdf=RangeControl
3555127
The utility distribution function for a single topic* is the normal distribution that maps utility to probability density for that topic. U is the utility, and p is the probability density. It's integrating over that curve with respect to utility. The aim is to find the utility contribution of a vote "yes" or "no" on a topic when the utility of "topic X is selected" is normally distributed and when the outcome is probabilistic.
* That's assuming that each topic for each voter has a separate normal distribution. I'm not sure if that's what's causing the confusion. I thought that's what you intended in 3550440.
Your point is still right, that this fails for the one-voter case and in all other cases where there is a tie. Minus possibly some missing assumptions, I still think 3555016 is right in cases without ties, but it causes more ties than are necessary. I'll post a more thorough response tomorrow night.
3555243 A single topic can only have a probability, not a probability density (unless there are an infinite number of topics), and not a probability density function.
What curve?
Oh, I get it. You're talking about the probability density function giving its utility; I was talking about the probability of it being chosen.
3555127
I take back what I said in 3555243. You're right that my analysis was flawed. In the correct analysis, voting for everything should be equivalent to voting for nothing, which was clearly not the case in mine.
I think you have to define the outcome in the case of a tie for this to work. You can say that we're ignoring the cases where there are ties, but then the strategy for the two-voter case is straightforward: just vote for your one topic with the highest utility. The other voter's options are to (1) vote with you, (2) don't vote, or (3) cause a tie. All of the non-ignored cases suggest that that is a winning strategy.
For three voters, you can break it down into cases:
(1) Both of the other voters have voted, and the other two voters share some number of common votes. The best strategy is to limit your vote to those common ones, and select the one with the highest rank. Actual utility values don't matter here.
(2) Both of the other voters have voted, and the other two share no common votes. The best strategy is to vote for the one with the highest rank. Actual utility values don't matter here.
(3) Only one other voters has voted so far. You have to either vote with the first voter or try to vote with the second voter (whose votes are unknown).
(4) None of the other voters have voted yet. I won't go into this case since the previous cases are enough to show that voting order matters, and that rankings are not sufficient to determine the best selection.
The best strategy for (3) requires choosing between (a) increasing the chance the second-voter will vote with you by selecting more topics that first-voter didn't select, (b) increasing the gains from second-voter voting with you by selecting fewer-but-higher-ranked topics that first-voter didn't select, and (c) voting with first-voter, which yields a guaranteed outcome but is limited to the topics he selected. The best strategy here requires multiplying utilities with probabilities and adding them together, which can't be done given only utility rankings.
I've voted for my n highest topics that are all higher-ranked than any of first-voter's topics. Should I vote for my next-highest one as well? Yes, if second-voter would otherwise vote for one of first-voter's topics. No, if second-voter would otherwise vote for one of my topics. Each of the two scenarios has an associated probability*expected_gain from voting for that next-highest topic. The rational decision would be made by adding those values together and comparing it to zero. The sign of that sum cannot be computed from rankings and probabilities alone.
Hey, I think I get to vote now! How about these three:
"The New Sincerity"
"Why it's Easy to Write for MLP"
"An appraisal-based ontology of fiction"
Personally I'm more interested in various auction types than voting, so if anyone wants to debate the merits of a sealed-bid second-price auction vs. open bid ascending price auction, I'm game. Or we could discuss the winner's curse and how auction winners are likely to have over-bid for their item.
3555127
Voting for topic n decreases utility if the other voter voted for a topic in 1 .. n-1 and also voted for topic n, wouldn't it? I don't see why voting for topic n would decrease utility if the other voter voted for a topic in n+1 .. N. No topic in n+1 .. N can get selected since there is no way it can get a majority of votes.
3561866 3550106 3546912 That's 3 people, 6 votes, and me already 1 week behind:
"Tics, Stereotypes, and Character": A short, inconclusive piece musing whether a "round" or "deep" character is anything more than a sufficiently large pile of cheap tricks.
"The Creativity Revolution"
"The Mystery of Mysteries"
"The New Sincerity"
"Why it's Easy to Write for MLP"
"An appraisal-based ontology of fiction"