There’s a question in OkCupid that asks “In some sense, wouldn’t nuclear war be exciting?” which [I immediately answered no and rated everyone who said yes as completely undateable] I think falls into this same class of bug, but I can’t quite put my finger on how to describe it.
“In some sense, wouldn’t nuclear war be exciting?”
In some sense, it probably would, it’s just a sense that doesn’t have any weight to speak of in deciding whether a nuclear war is a good idea. Even reliably settled arguments are not one-sided; there are usually considerations aligned against even the most obviously right decisions, and denying the existence or correctness of such considerations damages one’s epistemic rationality.
I agree, but I’m really confused about the how the creators of that website intended for the question to help in deciding whether a user should date a certain individual. I wouldn’t be able to tell if they answered “yes” because “Yay! Explosions!” and completely disregarded human deaths, or if they were saying “Indeed, there exists a sense in which nuclear war would create more excitement than a lack of one.”
I feel like “excitement” carries a positive connotation, particularly in American culture, which makes me uneasy about any “yes” answers. :(
I’m really confused about the how the creators of that website intended for the question to help in deciding whether a user should date a certain individual. I wouldn’t be able to tell if they answered “yes” because [...] or if they were saying [...]
I think this is a stock trick personality tests use: give test-takers a question where the denotation & connotation conflict; see how each test-taker resolves the conflict; people who resolve it in the same way (i.e. give the same answer) are presumably more similar in personality/weltanschauung than people who resolve it differently.
This seemed obvious to me. The problem is the lack of “meta” options; where’s the hidden checkbox for people who saw all six possible chains of reasoning, analyzed each of them, have probabilistic answers on four of those, along with an objection to the premises of the fifth and want to scream at the sixth for its stupidity? (bogus example)
Some of us don’t like limiting ourselves to only one possible interpretation of a statement or question. Some of us consider at least four different interpretations by default as a matter of convenience, and only then afterwards settle on the one most likely to have been “intended” within context.
This behavior is the one I prefer, not the behavior of automatically resolving to one specific preferred interpretation without noticing the others. The particular example question, like so many others on that site, provides no means of distinguishing between these behaviors, other than a very time-consuming reading of all the comments (which also requires time investment from the question-answerer by writing a comment in explanation, but this in turn requires a specific response, which partly defeats the point of going meta).
Other websites sometimes sidestep the issue entirely by first testing for traits that have these effects and often outright rejecting those (potential members) that would “question” the questions, thereby pre-filtering members for compatibility with their testing methodologies.
It seems to me that people who argue with questionnaires might have a good bit in common with each other, and likewise for people who don’t argue with questionnaires.
A crude approach would be to just match up people by the number of questions they argue with and the amount they write. It would be more sophisticated to just let people see each other’s comments on the questionnaire.
where’s the hidden checkbox for people who saw all six possible chains of reasoning, analyzed each of them, have probabilistic answers on four of those, along with an objection to the premises of the fifth and want to scream at the sixth for its stupidity?
You picked the answer among the first four for which your probabilistic answer is highest, mark all the first four answers as acceptable in a potential date, and explain your reasoning in the comment section.
Oh, no, I think you misunderstand what parts of the question-problem I was talking about. To better characterize the bogus example, let’s flesh it out a bit:
Q: Which is healthier? ( ) Bleggs ( ) Rubes ( ) Both ( ) Neither
Now obviously, the first four chains of reasoning go as follow:
In most specific cases, presented an arbitrary thought-experiment-style choice of being handed a blegg or a rube, neither is good. Owning either a Blegg or a Rube will make you less physically healthy. So among the four options, “Neither” is clearly better. This is pretty certain, though some crack scientists do claim conclusive evidence that owning both at the same time can be healthy. But I don’t put much faith in their suspicious results.
“But!”, screams the more logically-minded, “the question isn’t about which of the four choices presented is better—it’s clear that the fourth option is intended to mean ‘neither bleggs nor rubes are healthier’, not that you should pick neither. So the thought experiment implied means you have to pick one of the two, and in that case Bleggs are clearly marginally better!” Okay, fine. So Bleggs are most likely healthier if you have to choose one of the two—they’re unlikely to be equally unhealthy or healthy, after all.
But let’s take a step back for a moment. If you look at the grand scheme of things, at a macro scale, Rubes do reduce the total amount of Bleggs and Rubes, because each Rube will destroy at least five Bleggs. So in the grand scheme of things, having Rubes is healthier than having Bleggs, if we can’t attack the source! Clearly, both of the previous chains of reasoning are too narrow-minded and don’t think of the big picture. On a large scale, the Rubes are indeed healthier-per-unit than the Bleggs. Probably.
Ah, but what if it is implied that this is an all-or-nothing paradigm, and what if others interpret it this way? Then, obviously, the complete absence of both Bleggs and Rubes would be a Very Bad Thing™, since we require Bleggs and Rubes to produce Tormogluts, a necessary component of modern human prosperity! Thus, both are (probably) healthier than only having one or the other (and obviously better than neither).
...
On the other hand, Bleggs and Rubes are unnatural, unsustainable in the long term, and we will soon need to research new ways to produce Tormogluts. Most people who see you advocating for them will automatically match you as The Enemy, so you should pick “Neither”, even though that’s not what the question implies. But this is a shitty situation, and if someone reading my answer to this question interprets it this way, I don’t care to befriend them anyway. So I reject this answer.
And let’s not even think of what the Kurgle fanatics have to say about this question. The horror.
Assuming all of the above went through your mind in a few seconds very rapidly when you first read the question… what answer do you choose? Do you also put a preference filter for other people’s answers? Just choosing the higher or most confident probability from the above isn’t going to cut it if this question matters to you a lot.
I used to pick the “least bad” answer in such cases, but then I decided to clear all my previous answers, and now when I see a question to which the answer I wish I could give is “Mu” or “ADBOC” or “Taboo $word” or “Avada Ked--[oh right, new censorship policy, sorry]”, I just skip it.
“Avada Ked--[oh right, new censorship policy, sorry]”
looks baffled
Avada Kedavra is a spell trigger for a death spell. In the context implies the urge to a respond to a stupid, misleading or perhaps disingenuous question with the application of power, violence or killing rather than compliance with the form of the question. Since it isn’t specific or in reference to any actual people this doesn’t technically violate the new censorship policy but it is still close enough that I laughed at Army’s joke!
it isn’t specific or in reference to any actual people
I’d guess that “people who write match questions on OkCupid” is specific enough that actually advocating violence against them would be against the spirit of the policy. (I can’t imagine someone actually doing that, or imagine someone imagining someone doing that, but I’ve lost count of the times I’ve underestimated the validity of Poe’s Law.)
Yep, that’s the reasoning I followed in the earlier comment. A person who saw all six possible chains would decide the question wasn’t useful and would refuse to answer it, hopefully. ^_^
I’m really confused about the how the creators of that website intended for the question to help in deciding whether a user should date a certain individual.
It’s up to the users: you have to both provide your own answer and decide which answers you would consider acceptable in a potential match (and specify how much of a big deal would it be for a potential match to pick a different answer). If you want to provide your answer but don’t want to discriminate potential matches based on their answers, you can mark all possible answers as acceptable or equivalently mark the question as irrelevant. (And many of the questions are written by users of the site, rather than by its creators; I don’t remember whether the one about nuclear war is.)
The matching algorithm is described here. (Its unBayesianity makes me cringe—the rarer a particular answer to a particular question is, the larger the effect of someone picking that answer ought to be—but still.)
I wish you could condition on whether the user ignored the question or not, but I don’t think you can. Also, I’m pretty sure the nuclear war one wasn’t a user question.
I feel like “excitement” carries a positive connotation … which makes me uneasy about any “yes” answers.
That’s the point: you should (in particular) be comfortable with entertaining arguments for horrible things that carry positive connotations (just don’t get carried away :-). The correctness of these arguments won’t in general depend on whether their connotations align with those of the decisions reached upon considering all relevant arguments.
You’re saying a lot of technically correct things that don’t seem to be engaging what I’m saying. =/
Yes, I agree that there is some value in entertaining a “yes, during a nuclear war, there will be may be some more (positively) exciting things than in peacetime.” This is something to take into account when deciding whether or not you should go to nuclear war.
Meanwhile, if you’re trying to use the question to gauge the moral compass of the answering person, the “nuclear war is great and fun thing!” answer is not readily distinguishable from the “I am carefully entertaining the argument for a horrible thing” answer. Which is similar to the way “awesomeness” sometimes leads to … awesome starvation schemes?
Still, in questions which have one correct answer (e.g. “Which is bigger, the earth or the sun” or “STALE is to STEAL as 89475 is to...”), I only mark the correct answer and “I don’t know” (if it’s there) as acceptable, and I mark the question as “Mandatory” if the “I don’t know” answer is available. It’s OK to be ignorant, but it’s not OK to not admit it.
That’s not even the worst one: “Which is worse: starving children or abused animals?” with possible answers “Starving children”, “Abused animals”, “Neither, both are good” and “Neither, they are equally bad.” I’m curious to know whether there’s actually somebody who picks “Neither, both are good.”
Making jokes in data that’s going to have statistics done on by computers is one of my pet peeves. For example, I’ve seen obviously sarcastic book reviews on Amazon.com where the number of stars matched the letter of the review rather than its spirit. (And IIRC, Yvain mentioned how people answered stuff like “over 9000” in the LW survey.)
Oh yeah! I remember that one now that you mention it! I wish I knew how they came up with these.
I think, personally, I would answer “Both are bad, but I’m not going to bother quantifying which is worse until I am confronted with an actual situation in which I have to.” Which is definely not the same as “both are equally bad.” Bah!
“There is nothing so exhilarating in all the world as being shot at with no result.”
Attributed, in various forms, to Winston Churchill. What war is, is intense. Soldiers who have seen combat duty often miss it in peacetime, or in civilian life. In Britain after the second World War, even many civilians found the peace a bit of a let down.
So yes, in a very ordinary sense, nuclear war would be exciting, especially if you survive it.
Imagine you have the superpower: Movie Hero. You are guaranteed to escape from all situations, however dire, and whatever extremity of privation and suffering you may have to go through (but you will have to go through it) in the process of clawing your way to the Happy Ending. You also get a chance to play a pivotal role in whatever world crisis forces itself on you. How would you then feel about seeing the world slide towards imminent nuclear war?
Unfortunately, I think I would prefer not to go on dates with people who spend too much time imagining that they are Movie Heroes. A little bit is okay! =P
Wouldn’t the failure to acknowledge all the excitement nuclear war would cause be an example of the horns effect?
I immediately answered no and rated everyone who said yes as completely undateable
I can understand answering no for emotional or political reasons, but rating the epistemically correct answer as undateable? That’s… a good reason for me to answer such questions honestly, actually.
So they have a mechanism for you to write an explanatory comment, right? But they don’t allow you to filter on the existence of an explanatory comment, which would allow someone to explain their thought process—which I think is really necessary because “exciting” does strongly connote “good idea” the way “awesome” does. In which case, I would expect a person trying to avoid the horns effect to just refuse to answer the question on the grounds that it’s misleading as a moral compass gauge because answering “yes” might cause them to get filtered out because you can’t condition on the existence of an explanation. So I expected most “yes” answers to be generally unaware people. I don’t think that question was intended for rational arguments for or against nuclear war; I think it was intended for … morality. I admit “completely undateable” is an exaggeration, but I think I decided engaging that question was a red flag for immaturity.
But that’s why I’m really confused why that question was there in the first place because it doesn’t distinguish those two groups of people—the ones that are thinking really really carefully and the ones that aren’t thinking at all. It’s bad for morality!
Well, it might be exciting for people fighting it from within their anti-radiation shelters, but it’d be such a drag for people having their friends and family killed. The total excitingness I’m pretty sure would be negative.
That’s why I thought it was a useless question. Because it’s not asking for the overall total excitingness of nuclear war. It’s asking if there exists a type of excitingness that nuclear war does have a net positive in. Which, probably yes, but it’s so not very relevant to anything. =P
There’s a question in OkCupid that asks “In some sense, wouldn’t nuclear war be exciting?” which [I immediately answered no and rated everyone who said yes as completely undateable] I think falls into this same class of bug, but I can’t quite put my finger on how to describe it.
In some sense, it probably would, it’s just a sense that doesn’t have any weight to speak of in deciding whether a nuclear war is a good idea. Even reliably settled arguments are not one-sided; there are usually considerations aligned against even the most obviously right decisions, and denying the existence or correctness of such considerations damages one’s epistemic rationality.
I agree, but I’m really confused about the how the creators of that website intended for the question to help in deciding whether a user should date a certain individual. I wouldn’t be able to tell if they answered “yes” because “Yay! Explosions!” and completely disregarded human deaths, or if they were saying “Indeed, there exists a sense in which nuclear war would create more excitement than a lack of one.”
I feel like “excitement” carries a positive connotation, particularly in American culture, which makes me uneasy about any “yes” answers. :(
I think this is a stock trick personality tests use: give test-takers a question where the denotation & connotation conflict; see how each test-taker resolves the conflict; people who resolve it in the same way (i.e. give the same answer) are presumably more similar in personality/weltanschauung than people who resolve it differently.
This seemed obvious to me. The problem is the lack of “meta” options; where’s the hidden checkbox for people who saw all six possible chains of reasoning, analyzed each of them, have probabilistic answers on four of those, along with an objection to the premises of the fifth and want to scream at the sixth for its stupidity? (bogus example)
Some of us don’t like limiting ourselves to only one possible interpretation of a statement or question. Some of us consider at least four different interpretations by default as a matter of convenience, and only then afterwards settle on the one most likely to have been “intended” within context.
This behavior is the one I prefer, not the behavior of automatically resolving to one specific preferred interpretation without noticing the others. The particular example question, like so many others on that site, provides no means of distinguishing between these behaviors, other than a very time-consuming reading of all the comments (which also requires time investment from the question-answerer by writing a comment in explanation, but this in turn requires a specific response, which partly defeats the point of going meta).
Other websites sometimes sidestep the issue entirely by first testing for traits that have these effects and often outright rejecting those (potential members) that would “question” the questions, thereby pre-filtering members for compatibility with their testing methodologies.
It seems to me that people who argue with questionnaires might have a good bit in common with each other, and likewise for people who don’t argue with questionnaires.
A crude approach would be to just match up people by the number of questions they argue with and the amount they write. It would be more sophisticated to just let people see each other’s comments on the questionnaire.
You picked the answer among the first four for which your probabilistic answer is highest, mark all the first four answers as acceptable in a potential date, and explain your reasoning in the comment section.
Oh, no, I think you misunderstand what parts of the question-problem I was talking about. To better characterize the bogus example, let’s flesh it out a bit:
Q: Which is healthier?
( ) Bleggs
( ) Rubes
( ) Both
( ) Neither
Now obviously, the first four chains of reasoning go as follow:
In most specific cases, presented an arbitrary thought-experiment-style choice of being handed a blegg or a rube, neither is good. Owning either a Blegg or a Rube will make you less physically healthy. So among the four options, “Neither” is clearly better. This is pretty certain, though some crack scientists do claim conclusive evidence that owning both at the same time can be healthy. But I don’t put much faith in their suspicious results.
“But!”, screams the more logically-minded, “the question isn’t about which of the four choices presented is better—it’s clear that the fourth option is intended to mean ‘neither bleggs nor rubes are healthier’, not that you should pick neither. So the thought experiment implied means you have to pick one of the two, and in that case Bleggs are clearly marginally better!” Okay, fine. So Bleggs are most likely healthier if you have to choose one of the two—they’re unlikely to be equally unhealthy or healthy, after all.
But let’s take a step back for a moment. If you look at the grand scheme of things, at a macro scale, Rubes do reduce the total amount of Bleggs and Rubes, because each Rube will destroy at least five Bleggs. So in the grand scheme of things, having Rubes is healthier than having Bleggs, if we can’t attack the source! Clearly, both of the previous chains of reasoning are too narrow-minded and don’t think of the big picture. On a large scale, the Rubes are indeed healthier-per-unit than the Bleggs. Probably.
Ah, but what if it is implied that this is an all-or-nothing paradigm, and what if others interpret it this way? Then, obviously, the complete absence of both Bleggs and Rubes would be a Very Bad Thing™, since we require Bleggs and Rubes to produce Tormogluts, a necessary component of modern human prosperity! Thus, both are (probably) healthier than only having one or the other (and obviously better than neither).
...
On the other hand, Bleggs and Rubes are unnatural, unsustainable in the long term, and we will soon need to research new ways to produce Tormogluts. Most people who see you advocating for them will automatically match you as The Enemy, so you should pick “Neither”, even though that’s not what the question implies. But this is a shitty situation, and if someone reading my answer to this question interprets it this way, I don’t care to befriend them anyway. So I reject this answer.
And let’s not even think of what the Kurgle fanatics have to say about this question. The horror.
Assuming all of the above went through your mind in a few seconds very rapidly when you first read the question… what answer do you choose? Do you also put a preference filter for other people’s answers? Just choosing the higher or most confident probability from the above isn’t going to cut it if this question matters to you a lot.
I used to pick the “least bad” answer in such cases, but then I decided to clear all my previous answers, and now when I see a question to which the answer I wish I could give is “Mu” or “ADBOC” or “Taboo $word” or “Avada Ked--[oh right, new censorship policy, sorry]”, I just skip it.
Good idea, although it’s still useful to be able to choose in case you need to fill in all the questions.
looks baffled
Avada Kedavra is a spell trigger for a death spell. In the context implies the urge to a respond to a stupid, misleading or perhaps disingenuous question with the application of power, violence or killing rather than compliance with the form of the question. Since it isn’t specific or in reference to any actual people this doesn’t technically violate the new censorship policy but it is still close enough that I laughed at Army’s joke!
I’d guess that “people who write match questions on OkCupid” is specific enough that actually advocating violence against them would be against the spirit of the policy. (I can’t imagine someone actually doing that, or imagine someone imagining someone doing that, but I’ve lost count of the times I’ve underestimated the validity of Poe’s Law.)
Ah, right. That is pretty funny, actually.
Yep, that’s the reasoning I followed in the earlier comment. A person who saw all six possible chains would decide the question wasn’t useful and would refuse to answer it, hopefully. ^_^
It’s up to the users: you have to both provide your own answer and decide which answers you would consider acceptable in a potential match (and specify how much of a big deal would it be for a potential match to pick a different answer). If you want to provide your answer but don’t want to discriminate potential matches based on their answers, you can mark all possible answers as acceptable or equivalently mark the question as irrelevant. (And many of the questions are written by users of the site, rather than by its creators; I don’t remember whether the one about nuclear war is.)
The matching algorithm is described here. (Its unBayesianity makes me cringe—the rarer a particular answer to a particular question is, the larger the effect of someone picking that answer ought to be—but still.)
I wish you could condition on whether the user ignored the question or not, but I don’t think you can. Also, I’m pretty sure the nuclear war one wasn’t a user question.
That’s the point: you should (in particular) be comfortable with entertaining arguments for horrible things that carry positive connotations (just don’t get carried away :-). The correctness of these arguments won’t in general depend on whether their connotations align with those of the decisions reached upon considering all relevant arguments.
You’re saying a lot of technically correct things that don’t seem to be engaging what I’m saying. =/
Yes, I agree that there is some value in entertaining a “yes, during a nuclear war, there will be may be some more (positively) exciting things than in peacetime.” This is something to take into account when deciding whether or not you should go to nuclear war.
Meanwhile, if you’re trying to use the question to gauge the moral compass of the answering person, the “nuclear war is great and fun thing!” answer is not readily distinguishable from the “I am carefully entertaining the argument for a horrible thing” answer. Which is similar to the way “awesomeness” sometimes leads to … awesome starvation schemes?
I think that using factual beliefs to signal something other than knowledge about the world is a bad idea. It encourages lying to yourself and others.
Still, in questions which have one correct answer (e.g. “Which is bigger, the earth or the sun” or “STALE is to STEAL as 89475 is to...”), I only mark the correct answer and “I don’t know” (if it’s there) as acceptable, and I mark the question as “Mandatory” if the “I don’t know” answer is available. It’s OK to be ignorant, but it’s not OK to not admit it.
That’s not even the worst one: “Which is worse: starving children or abused animals?” with possible answers “Starving children”, “Abused animals”, “Neither, both are good” and “Neither, they are equally bad.” I’m curious to know whether there’s actually somebody who picks “Neither, both are good.”
I saw someone who answered that way, but it must have been as a joke. Not a good thing to do for matching purposes...
Making jokes in data that’s going to have statistics done on by computers is one of my pet peeves. For example, I’ve seen obviously sarcastic book reviews on Amazon.com where the number of stars matched the letter of the review rather than its spirit. (And IIRC, Yvain mentioned how people answered stuff like “over 9000” in the LW survey.)
Oh yeah! I remember that one now that you mention it! I wish I knew how they came up with these.
I think, personally, I would answer “Both are bad, but I’m not going to bother quantifying which is worse until I am confronted with an actual situation in which I have to.” Which is definely not the same as “both are equally bad.” Bah!
“There is nothing so exhilarating in all the world as being shot at with no result.”
Attributed, in various forms, to Winston Churchill. What war is, is intense. Soldiers who have seen combat duty often miss it in peacetime, or in civilian life. In Britain after the second World War, even many civilians found the peace a bit of a let down.
So yes, in a very ordinary sense, nuclear war would be exciting, especially if you survive it.
Imagine you have the superpower: Movie Hero. You are guaranteed to escape from all situations, however dire, and whatever extremity of privation and suffering you may have to go through (but you will have to go through it) in the process of clawing your way to the Happy Ending. You also get a chance to play a pivotal role in whatever world crisis forces itself on you. How would you then feel about seeing the world slide towards imminent nuclear war?
Pretty miserable.
Unfortunately, I think I would prefer not to go on dates with people who spend too much time imagining that they are Movie Heroes. A little bit is okay! =P
Wouldn’t the failure to acknowledge all the excitement nuclear war would cause be an example of the horns effect?
I can understand answering no for emotional or political reasons, but rating the epistemically correct answer as undateable? That’s… a good reason for me to answer such questions honestly, actually.
So they have a mechanism for you to write an explanatory comment, right? But they don’t allow you to filter on the existence of an explanatory comment, which would allow someone to explain their thought process—which I think is really necessary because “exciting” does strongly connote “good idea” the way “awesome” does. In which case, I would expect a person trying to avoid the horns effect to just refuse to answer the question on the grounds that it’s misleading as a moral compass gauge because answering “yes” might cause them to get filtered out because you can’t condition on the existence of an explanation. So I expected most “yes” answers to be generally unaware people. I don’t think that question was intended for rational arguments for or against nuclear war; I think it was intended for … morality. I admit “completely undateable” is an exaggeration, but I think I decided engaging that question was a red flag for immaturity.
But that’s why I’m really confused why that question was there in the first place because it doesn’t distinguish those two groups of people—the ones that are thinking really really carefully and the ones that aren’t thinking at all. It’s bad for morality!
What senses of ‘exciting’ do you think exist? Why wouldn’t you date someone who thinks that interesting times (in the Chinese sense) are exciting?
What’s the Chinese sense? o.O
This is a reference to an alleged Chinese proverb.
Well, it might be exciting for people fighting it from within their anti-radiation shelters, but it’d be such a drag for people having their friends and family killed. The total excitingness I’m pretty sure would be negative.
That’s why I thought it was a useless question. Because it’s not asking for the overall total excitingness of nuclear war. It’s asking if there exists a type of excitingness that nuclear war does have a net positive in. Which, probably yes, but it’s so not very relevant to anything. =P