Following your heart is another tolerable option
June 21, 2014 5:16 AM   Subscribe

 
Fun column, but Krulwich is wrong when he says "the Very Best Candidate might show up in that first 36.8 percent — in which case you'll be stuck with second best." If the best candidate is in the initial group passed over, you end up with the very last candidate, who could be anywhere from second best to worst.
posted by grimmelm at 5:51 AM on June 21, 2014 [17 favorites]


I knew I'd married the right woman because she told me. 15 years later, she's still right (so I'm told).
posted by arcticseal at 5:56 AM on June 21, 2014 [16 favorites]


The implied priorities are telling here. Physical attractiveness is a screening procedure, after which personality and upbringing (treated as interchangeable by Kepler) are the second-order preferences. And the whole thing fits into an "employer-job candidate" dynamic that almost no one would now mistake for friendly respect, let alone romantic love.

However, I'm not sure that the idea of "following your heart" is any less objectifying in the end. It still makes it possible to turn relationships with other people into a utility problem -- who best enables me to feel as I want to? -- and that doesn't seem likely to me to reflect the reality of either Kepler's marriage or most relationships. It is, of course, a much better starting point.

Notice, though, that "their interest in me" becomes an accidental factor, something that only emerges if it means the "candidate" removes herself. And even then, it's framed as an incompatible personality factor -- "not the waiting type," rather than "not going to wait for *you*" or "hey, are you making some kind of quantitative list and running through all the options instead of relating to me like I'm a non-interchangeable person?"
posted by kewb at 5:58 AM on June 21, 2014 [12 favorites]


You know that the article is about the formula, right? Kepler's personal preferences are incidental (we all have preferences)
posted by sparklemotion at 6:10 AM on June 21, 2014 [5 favorites]


You know that the article is about the formula, right? Kepler's personal preferences are incidental (we all have preferences)

The article's about applying game theory to personal preferences in order to optimize lifetime happiness with a romantic relationship, only to then turn and say that whimsy and serendipity can still beat rational process.

None of that lets you actually examine the preferences, precisely because it assumes that all preferences are interchangeable and of equal value. This implies that preference is idiosyncratic but morally and socially neutral, which is, of course, not true.

If you're not examining the preferences, you're enabling some pretty horrible things, in the same way that the Prisoner's Dilemma is great for figuring out what best gets the two prisoners what they want, but really bad at helping you figure out if what they want is actually any good for them or anyone else. (Well, really, the whole "preferences are incidental" idea comes from a place that's deeply suspicious of trying to use anything other than personal desires and rationalized allocations to define "good," but that's a bait-shop shelf full of canned worms.)
posted by kewb at 6:18 AM on June 21, 2014 [12 favorites]


When you have a constrained list (like Keppler's ladies), this might make sense. But what about someone who is online dating in a big city, with a seemingly endless list of potential dates? How do they know when they have hit 36.8 percent of the total?

The overall advice to satisfice rather than optimize in a situation with imperfect information and constraints is good, though, and I've seen friends make unfortunate dating choices by not following this.
posted by Dip Flash at 6:23 AM on June 21, 2014 [2 favorites]


Sounds like a bunch of woo to me.
posted by drlith at 6:32 AM on June 21, 2014 [18 favorites]


He totally missed the "Let my best friend set me up on a blind date with a girl he is breaking a date with" methodology. In my experience it is 100% successful. Did it once, will be married 23 years in August.
posted by COD at 6:36 AM on June 21, 2014 [18 favorites]


I can't find the article now, but I swear I've read something by a psychologist about decision making that struck upon a very similar answer for entirely different reasons. If you have complex criteria, and a large pool of candidates (houses, phones, restaurants, dates...) which are all "good enough", looking closely at too many of them can be counter-productive. You probably won't find one candidate that's obviously better than all the others. But whatever you choose in the end, you'll definitely remember the unique highlights in candidates #3, #7, and #12 that you gave up, and that most of the candidates didn't have the specific minor flaw that you settled with.

So the advice was: do a brief survey of the field, and pick out e.g. 5-6 that you rate as "good enough". Then start looking again, and choose the first one you find that beats your baseline group. The one you choose probably isn't the "best" by all of your starting criteria, but you weren't that likely to find an objective "best" anyway, and this way you'll almost certainly be happier about the decision.

grimmelm - If the best candidate is in the initial group passed over, you end up with the very last candidate, who could be anywhere from second best to worst.

Tim Minchin - If I didn't have you
posted by metaBugs at 6:47 AM on June 21, 2014 [10 favorites]


Thinking on this..

This can be construed as a combination of a sort of search problem, for the maximal lists of matching people of the preferred gender who score high on a good scoring algorithm. One that takes similarity of personality, physical appearance, etc etc into account. Personality apparently accounts for 30% of variance, although it's actually simpler than simple similarity when it comes to Big Five personality: it's just that conscientious people are better off with respect to marital satisfaction, and neurotic people are worse off. Of course, it's one of those factor analytic things that the psychologists pull out and give fanciful names to eigenvectors, so you don't know.

I think that most of the companies out there go with an ensemble of a bunch of features which are the answers to a bunch of questions, do a dimensionality reduction and get their predictions from there, so it's less principled but probably as effective.

Of course, this is one of those low-recall-is-actually-fine problems, counter to the implicit claims of most romances where You Must Find Your True Love And There Is One True Love For You Which Presumably Means That You Must Do A Complete Search. I mean, Google Search is fine with 0.05% recall: they only show you like 500 results, tops, despite the fact that they really do get like 500,000,000 pages for each search. Unlike Google Search, I suspect PageRank is not a good feature (no real meaning to it in this problem domain). If you limit n a lot, O(n^2) is still alright. And you certainly don't need to look at 36.8% of people in a large city.

Features is definitely important in this one for the weighing algorithm. MHC compatibility is probably interesting, and also O(n^2) for calculation, iff we get a better objective measure of MHC compatibility than "sniff this dude's t-shirt, young lady, and see how attractive the smell is". I'm sure there is one somewhere, I'm just not familiar with the literature.

There seems to be a heritability in marital satisfactions and coping and attributional style, which I suspect they can't do twin studies too well with and which I suspect, therefore, can be explained away by similar backgrounds. Coping style, however, is probably a feature.

If you remember the post some time ago, Gottman has his 94% predictive model on whether a marriage will last. However, these are features specific to the marriage itself, and therefore they wouldn't be that good for a matching algorithm. Much of the rest of the prediction-of-marital-satisfaction literature works on this, also: structure of how they deal with their children and their fights, life stressors and transitions. I also think that the physiological studies would probably be less feasible.

One implicit problem with this is that it's an acceptance of the idea that intervention in these matters is not possible by third party people.
posted by curuinor at 6:48 AM on June 21, 2014 [2 favorites]


metaBugs - this is the work on satisficing by the great Herbert Simon, who used the idea to criticize economics. Most economic models assume nobody ever satisfices in any way, shape or form, depressingly.
posted by curuinor at 6:50 AM on June 21, 2014 [4 favorites]


Tangent: The subject brought to mind Deborah Kerr singing "Getting to Know You" from The King and I.
posted by Daddy-O at 6:57 AM on June 21, 2014 [1 favorite]


The obvious best solution is to interview them all and pick the best one. There needs to be some kind of constraint on the problem (time, etc) to make the one in the article the optimal solution. Given the stakes of a decision about marriage, it seems pretty obvious that spending the time to interview them all is ideal.
posted by empath at 7:11 AM on June 21, 2014


Further proof that engineers and people who think like this are a whole 'nother tribe speaking a different language from me.
posted by Pirate-Bartender-Zombie-Monkey at 7:19 AM on June 21, 2014 [5 favorites]


The obvious best solution is to interview them all and pick the best one. There needs to be some kind of constraint on the problem (time, etc) to make the one in the article the optimal solution. Given the stakes of a decision about marriage, it seems pretty obvious that spending the time to interview them all is ideal.

The constraint in the article is that one isn't able to go back and choose a candidate that they've already passed over. That seems pretty fair - the success rates must be abysmal for folks who go back to some partner from times past and say, "Oh, you know, I've had twelve relationships since we parted ways, and by now statistics suggest that I'm not going to do better than you. Let's get hitched."
posted by vathek at 7:34 AM on June 21, 2014 [9 favorites]


Why would you want to marry a girl? Surely that's illegal. Girls aren't old enoug to marry.

Gross, but I guess it could be worse. The article could have been 'How to Marry the Right Female.'
posted by winna at 7:45 AM on June 21, 2014 [20 favorites]


I let her pick me. Worked pretty well.
posted by jscalzi at 7:48 AM on June 21, 2014 [8 favorites]


Now let's assume there are infinite members of the desired gender. What then?
posted by blue_beetle at 7:51 AM on June 21, 2014 [1 favorite]


The whole courting / marriage thing is a framing device for discussing a statistical formula. I don't think anyone's proposing (heh) that real human beings should really approach real marriage this way for real.
posted by escape from the potato planet at 8:18 AM on June 21, 2014 [1 favorite]


curuinor - I immediately thought of Herbert Simon and satisficing, too.

I'd imagine that as long as your standards are high enough, there's not very much difference between satisficing and maximizing, and for something like choosing a life partner you should have high standards. The way it's presented in Kepler's dilemma sounds cynical, because he's judging on some shallow requirements and he doesn't allow the option of not choosing anybody. But if you find someone you truly want to marry, it's just as cynical to hold off because you're not sure they are the absolute best.
posted by Metroid Baby at 8:23 AM on June 21, 2014 [1 favorite]


I just kept making mistakes until I finally made the correct mistake.
posted by srboisvert at 8:25 AM on June 21, 2014 [6 favorites]


When you have a constrained list (like Keppler's ladies), this might make sense. But what about someone who is online dating in a big city, with a seemingly endless list of potential dates? How do they know when they have hit 36.8 percent of the total?

You'll never date everyone in the city, so constrain your pool to the number of people you expect to date. Say you're 30 and want to marry by 35. You feel that you can date on average 5 people per year. So using this strategy you date 9 people, then marry the next person you like more than any of those 9.
posted by justkevin at 8:45 AM on June 21, 2014 [3 favorites]


Two things I have no patience for before even getting past the first paragraph: "girl", also the wildly incorrect use of the m-dash.
posted by Sara C. at 8:48 AM on June 21, 2014 [4 favorites]


I learned about this in a math class, where it was called the "Motel Problem" - you are driving down a highway at night, looking for a place to stay, and you don't want to turn around once you've checked a place out. I prefer that formulation to the "Marriage" or "Secretary" formulations, which are fraught with sexism and human complexity.
posted by crazy_yeti at 8:58 AM on June 21, 2014 [13 favorites]


I think you guys are misinterpreting this problem. The point is not to maximize the quality of your choice. The goal is to maximize confidence in your choice. This requires some experience with evaluating the candidates.

I have heard folklore versions of this problem, like suggestions not to marry the first 6 women you date.
posted by charlie don't surf at 8:58 AM on June 21, 2014 [1 favorite]


I love to see science! support an anecdotal theory I've long held -- that romantic incompetence is an epiphenomenon of innumeracy, people who partner badly or not at all having no grasp for valid sample size, (if you will) significant digits in analysis of that sample, or the likely growth and decline curves of their varied attractive qualities.
posted by MattD at 8:59 AM on June 21, 2014 [4 favorites]


But what about someone who is online dating in a big city, with a seemingly endless list of potential dates? How do they know when they have hit 36.8 percent of the total?

I was thinking of this as I read the piece, as someone who, now in my 30s, has been on and off the dating market for far too long.

I think that, in the case of something that isn't as cut and dry as a finite list of job applicants, it's more worthwhile to look at the amount of time you've been doing the thing, and not the total possibilities of people you could theoretically date. There will always be more potential partners. There is, however, a finite amount of time a person has to meet these people. (Roughly mapped to the length of your life, but probably with other factors included, like childbearing years, availability of age appropriate partners, etc.)

So, I'm 33. I've been dating in a serious manner since I was 18. That's 15 years. I met the first person who I really felt like I could settle down with for life when I was 28. I think a decade of dating works pretty well for "first 36.8% of your adult life". At this point, being still single, I'm starting to feel like this is taking longer than I'd really like it to take, and that I'd rather pick someone and stop interviewing potential candidates. At this point in my life, if I met someone who was better than everyone I met in my first decade or so of dating, I would be confident enough to marry them.

However, there's a problem -- a problem that Kepler had as documented in the article -- not everyone you want to marry wants to marry you. This formula works well if it's a one-sided choice, like a job interview or an auto mechanic. But it really doesn't work for marriage at all, because both people have to choose each other. In fact, Kepler himself lost out on 3 candidates because he was too indecisive, which proves that the formula specifically doesn't work for marriage at all, to the extent that I'm not sure why Kepler's marriage was used as a frame for the piece.

(That said, I'm confused about how he ended up with wife #5, when the article listed her as one of the ones who declined to wait for him. Also, holy shit spouses must be dying right and left if wife #5 already had step-children to dote on.)
posted by Sara C. at 9:03 AM on June 21, 2014 [5 favorites]


I think another downside to this theory is it ignores that there are others out there engaged in a parallel search for potentially the same optimal solution.

In that setting, there's some value in settling earlier rather than later as the remaining candidates are increasingly those who have not been matched to someone else, increasing the risk that they are not desirable to a large class of people that includes you.

Plus, as others have pointed out, we're totally skipping over the other party's opinion on the player.
posted by pulposus at 9:10 AM on June 21, 2014 [4 favorites]


pulposus - these are more reasons why the "motel problem" formulation is preferable (the motel won't turn anyone away). This is supposed to be a math/statistics problem, not psychology/social science!
posted by crazy_yeti at 9:13 AM on June 21, 2014 [1 favorite]


crazy_yeti - I agree this is better for choosing a motel than a relationship, but it's a pity, because I've never needed help choosing a motel.
posted by pulposus at 9:22 AM on June 21, 2014


Sandman et al. have some more general remarks on the subject.
posted by Dr Dracator at 9:39 AM on June 21, 2014


The problem is that the women know what they're being interviewed for. It would be better to interview them without them knowing that they're being eliminated or accepted. This is really how the real world often works: people seem to pair off most quickly and effectively when they have a large pool of friends that they can get to know before choosing one and dating them.

If one isn't in that situation, then one could set up a series of auditions for something else and then pursue the people that seem worth pursuing. If one were an established film producer, that would make this quite easy. One could then take one's time, asking questions and evaluating candidates carefully, before making a choice.

Then again - now that I think of it, that was the plot of a movie, and she ended up being a serial killer who tortured people and kept them locked in a closet.

So maybe that wouldn't actually work.
posted by koeselitz at 9:52 AM on June 21, 2014


The 1/e formula comes from a mathematical analysis involving a fixed-length random series of numbers with an unknown probability distribution, and a (probabilistic) algorithm for identifying the maximum number in this series when no "backtracking" is allowed. There are so many reasons this does not apply to human situations, I fear I would fill up several pages if I got into it all: e.g. numbers unlike people can be lined up and ranked on a 1-dimensional axis, and with people you do have some a-priori sense of the (high-dimensional) probability distribution, unless you've never interacted with another human being. I think that calling this the "Marriage Problem", or "Motel Problem", is just a (misguided?) attempt to get non-mathematicians more interested in math - or perhaps just mathematicians being playful, to keep themselves and their students amused (math can get pretty dry). It's like: if you have a problem involving a circle and some chords, it sounds a lot more fun if you call it the "Birthday Cake Problem" or something like that. (And topology has the "Ham Sandwich Theorem" and the "Hairy Ball Theorem", but that's getting off-topic).
posted by crazy_yeti at 9:59 AM on June 21, 2014 [6 favorites]


Only works for a spherical cow in a vacuum
posted by Fists O'Fury at 10:04 AM on June 21, 2014 [9 favorites]


I've never needed help choosing a motel.

So what algorithm do you use?

(fwiw, the beauty/dowry/secretary/motel/maximum-item-in-sequence problem is related to a big topic in computing these days, stream processing and probabilistic data structures, for problems where the amounts of data to analyze is way too large to hold in memory. Counting the number of items in a very long sequence is easy, of course, as is calculating basic statistical metrics, but how do you find possible duplicates, or figure out the number of distinct values, or the ten most common ones, without storing every single value you've seen this far?)

(it's not that hard, actually, you just ask wizards to design the algorithms for you, or at least have them calculate the appropriate parameters for the algorithm you're using.)
posted by effbot at 10:04 AM on June 21, 2014 [7 favorites]


The problem is that the women know what they're being interviewed for.

So, you're saying you've never seen Audition, right?
posted by valkane at 10:48 AM on June 21, 2014


Sounds like a bunch of woo to me.
posted by drlith at 6:32 AM on June 21 [4 favorites +] [!]


Indeed. Pitching woo.
posted by chavenet at 10:52 AM on June 21, 2014 [3 favorites]


To be applicable for dating---or even house shopping---I can't help thinking this must need a correction of some kind for rejection rate, i.e., for how likely it is it won't work out after the seeker decides they like it and tries to pursue.

I think it's common in dating to have it feel good to one, but not the other. Maybe something like a 10% chance A likes B, 10% B likes A and 1% whoopee. Ok, maybe a bit lower, but that makes the math easier. It scales with B-likes-A. Doesn't the strategy change depending how desirable Kepler is to the people he's courting, since he may have to try wooing several people after he crosses the cut-off.

Things get muddled even more if pursuing someone "better than your station" makes good outcomes less likely, if someone you'd prefer more is more likely to reject you. There's a bit of truth to that now, though less so and less explicitly so than Kepler's time.
posted by spbmp at 10:54 AM on June 21, 2014


Only works for a spherical cow in a vacuum


Dammit.

I botched the application of that joke.

Obviously should have been:

Only works for a spherical girl in a vacuum

tl;dr: Dammit
posted by Fists O'Fury at 11:04 AM on June 21, 2014 [3 favorites]


A more complete model of marriage is the Stable Marriage Problem where you want to pair two classes of people (e.g, men and women) where each member has a list of preferences for the other. The "stable solution" is where no one would rather trade up to someone who would prefer them instead of their current partner. [Amazingly, the solution is only O(n^2) -- not combinatorial hard.]

Of course, real life is more complicated: their aren't two sets, we have incomplete knowledge of what we want, and what we want and our attractiveness to others changes daily.

[PS: A similar model is used to pair medical school graduates to residency programs.]
posted by eigenman at 11:06 AM on June 21, 2014 [2 favorites]


The first 36.8 times I saw this problem, FWIW, it was framed as a princess who has to marry before a deadline for dynastic reasons, and needs to choose among the N suitors lined up at the castle gates. The impersonal process seems a little less out of place in that situation (and no, I don't know why she can't meet them all and then go back and and marry her favorite).
posted by hattifattener at 1:37 PM on June 21, 2014 [2 favorites]


Flag the first 38% of new posts and move on, then comment in the first post that's better than those.
posted by michaelh at 3:54 PM on June 21, 2014 [2 favorites]


(That said, I'm confused about how he ended up with wife #5, when the article listed her as one of the ones who declined to wait for him. Also, holy shit spouses must be dying right and left if wife #5 already had step-children to dote on.)

Sara C. -- the most common cause of death for women was childbirth. In a world of no birth control childbirth was pretty unavoidable.

I assume that he made the best choice because he went after the one he actually liked. And I'm also gonna assume that he liked her in part because she wouldn't put up with waiting around to be picked like a prize goat.
posted by jrochest at 4:36 PM on June 21, 2014


I'm aware that women died in childbirth.

However, for a woman to already have stepchildren and be seeking another marriage, that would mean those kids' dad was also dead.

We've got three dead parents (Kepler's late wife, as well) to make one marriage. Which, wow.
posted by Sara C. at 4:53 PM on June 21, 2014


Ah, I get it. I missed that -- agreed, that's a high death toll.
posted by jrochest at 6:00 PM on June 21, 2014


Isn't this just ordinary statistical sampling, but with a p of 50%? I guess not, because you choose above the max value, not above the mean.

For a very large population, you could choose a reasonable sample size, say 30, and date that many without intending to marry. Assign each a satisfaction score. Calculate mean and standard deviation, assume that for the entire population, then choose anyone you date after that who scores above a certain z-score.

Didn't I read a thing in the last year about a guy who got busted doing essentially this and got the internet hate machine focused on him for a while?
posted by ctmf at 7:29 PM on June 21, 2014


From personal experience, my spouse and I are WOEFULLY mismatched. I mean, you name a compatibility test or other test of suitability and it will come back "These people should not spend time alone in the same room together." (Seriously, name a criterion and we don't match. Extravert/introvert. Mega family/micro family. Siblings/only child. East Coast/West Coast. Gets louder with enthusiasm YAY/ow my ears hurt. Metalhead/alternakid. Ask/Guess. Ate hamburgers literally every night from age 10 until moved out/earthy crunchy veggie lover. It goes on, and on, and on.)

Next Friday is our 16th wedding anniversary. October will mark 19 years together. We regularly look at each other and ask "What the hell are we doing together?" (Followed by "Don't answer that, because whatever it is, it's working.")

IMO, don't ignore red flags, but "compatibility" and matching on individual criteria (whatever they may be) are overrated.
posted by Lexica at 7:45 PM on June 21, 2014 [3 favorites]


Yeah, I think it's really hard to come up with any kind of appropriate mathematical model that will apply to something like deciding who to marry. Especially since, well, what's the ultimate test of success? "Person married me"? People enter into marriages for terrible reasons all the time, and marriages end.

(Not that you and your spouse are going to divorce, Lexica, it's just that compatibility, love, and marriage are huge concepts that aren't as binary as something needs to be in order to be a worthwhile mathematical model.)
posted by Sara C. at 7:51 PM on June 21, 2014


Further proof that engineers and people who think like this are a whole 'nother tribe speaking a different language from me.

Not all engineers think like this. I've met very few who do, actually. (Thank goodness.)
posted by heisenberg at 8:32 PM on June 21, 2014


Only works for a spherical cow in a vacuum
posted by Fists O'Fury at 1:04 PM on June 21 [8 favorites +] [!]


That's a hell of a way to talk about your fiancée.
posted by McCoy Pauley at 10:26 PM on June 21, 2014 [3 favorites]


Not all engineers think like this.

Many engineers recognize recreational mathematics problems, though (as do many non-engineers, even).

(Cannot help imagining Martin Gardner on Metafilter here. "Five men and a monkey are shipwrecked on a desert island." "Bechdel fail, dude." "A young man who lives on Manhattan has two girlfriends, one in Brooklyn, one in the Bronx." "Ask.mefi is over there. Also they should DTMFA." "An island is inhabited by two tribes. Members of one tribe always tell the truth, members of the other always lie." "META.")
posted by effbot at 1:56 AM on June 22, 2014 [11 favorites]


☑ better than a dog anyhow.
posted by sebastienbailard at 2:00 AM on June 22, 2014 [2 favorites]


How To Marry The Right Girl

I generally work on the principle that she's likely to be the one in the fancy dress standing at the front of the church. This is an effective rule of thumb and I end up marrying the bride's father only about one in ten times.
posted by Grangousier at 2:38 AM on June 22, 2014 [8 favorites]


(Cannot help imagining Martin Gardner on Metafilter here. "Five men and a monkey are shipwrecked on a desert island." "Bechdel fail, dude." "A young man who lives on Manhattan has two girlfriends, one in Brooklyn, one in the Bronx." "Ask.mefi is over there. Also they should DTMFA." "An island is inhabited by two tribes. Members of one tribe always tell the truth, members of the other always lie." "META.")

Sorry, there is no magical, separate sphere where the rest of the world falls away before the joyful clarity of pure abstraction at play or the workings of unalloyed rational choice based strictly on numeracy. C'mon, that's why recreational math and game theory involve *telling a story* in the first place.

Situate the math to emphasize its applicability, and the situation you construct becomes part of what you have to discuss. And it's not as if engineering and mathematics fields don't have lots of real-world, well-documented problems with being an often rather smug and insular boys' club.
posted by kewb at 4:28 AM on June 22, 2014 [3 favorites]


Thanks for confirming my hypothesis.
posted by effbot at 4:50 AM on June 22, 2014 [2 favorites]


Is mathematics a "magical, separate sphere where the rest of the world falls away before the joyful clarity of pure abstraction"?

Only if you're doing it right.
posted by crazy_yeti at 8:23 AM on June 22, 2014 [2 favorites]


Only if you're doing it right.

As pure mathematics expressed as symbolic logic, sure. Start claiming it applies to social functions without considering non-mathematical contexts and you get into other realms entirely.

If the narrative is an element of trivial to no importance for framing the problem or idea, then either you don't need it or you can easily modify it to escape unwanted connotations.

If it's non-trivial, you are no longer in a world of pure abstraction and must engage with non-mathematical contexts in which the default assumption is not that all problems are utility problems *nor* that all utility is easily and simply quantifiable.
posted by kewb at 8:59 AM on June 22, 2014 [2 favorites]


This sort of problem is just supposed to be "for fun" and I doubt anyone in their right mind would try to use this algorithm to pick a spouse! But kewb, I agree, it's extremely problematic (and repellent to me) when people apply stuff like game theory to real life. The narrative is just intended to make the math a little more fun (which I think should be permissible), please don't take it so seriously.
posted by crazy_yeti at 10:16 AM on June 22, 2014 [3 favorites]


(also see my previous comment about changing the narrative to "motel problem" to avoid unwanted connotations)
posted by crazy_yeti at 10:19 AM on June 22, 2014 [1 favorite]


The narrative is just intended to make the math a little more fun (which I think should be permissible), please don't take it so seriously.

That's fair enough; I guess I was reacting more to the one or two comments here where people really were trying to apply it directly to real life and not giving enough benefit-of-the-hamburger.
posted by kewb at 10:45 AM on June 22, 2014


The problem is that the actual article is 100% about the framing concept and not really about math at all.

I mean, even the "check out 40% of the total options and then just pick something" angle isn't really the real math of the math. It's not a guideline for how to choose from a large group of options, it's just a math application.

For example i just ordered a computer mouse from Amazon. After setting up the search parameters (prime eligible, under $25, mac compatible), I didn't choose option 4, I looked at the top 10-15 results and ultimately chose the first one. Even if you consider hotels, why would you choose the fourth hotel you consider?
posted by Sara C. at 2:05 PM on June 22, 2014 [1 favorite]


>This sort of problem is just supposed to be "for fun" and I doubt anyone in their right mind would try to use this algorithm to pick a spouse!

>I guess I was reacting more to the one or two comments here where people really were trying to apply it directly to real life and not giving enough benefit-of-the-hamburger.


You people clearly didn't go to an engineering school.

I used to hang out with engineering students AND with a couple of math majors, a circle where the 'secretary' problem was discussed and USED as a guide to living.

(In those days, it was more about dating rather than actual marriage; but there was a general agreement that the odds of breaking up with a real-live-partner and then doing BETTER with the next (hypothetical) partner clearly must follow some sort of exponential decay.)

Decry it if you want to; but plenty of people try to evaluate that very situation ALL THE TIME - but try to do so intuitively, without availing themselves of the tools of mathematical analysis.

I don't get how "considering one's life prospects rationally" is such a terrible thing to do.
posted by AsYouKnow Bob at 2:24 PM on June 22, 2014 [2 favorites]


This sort of problem is just supposed to be "for fun"

I'm a woman who worked as an engineer for about a decade. The issue I have with this article is that it was really, really NOT fun for me, even though I find the underlying concepts really interesting. But from the "girl" in the title to the illustrations (seriously, wtf is up with the "athletic" one who totally isn't) it is overwhelmingly tailored to a male audience.

Between that and the engineers-think-like-X-normals-think-like-Y framing that is so depressingly common, where Y involves having social skills and interest in people as people, this really hit a raw nerve for me. It exemplifies, IMO, many of the reasons so few girls and women see the STEM fields as a viable path for themselves.
posted by heisenberg at 3:55 PM on June 22, 2014 [11 favorites]


I'm a woman who worked as an engineer for about a decade. I found this article to be a fun way to introducing an algorithm for satisficing that I hadn't heard of before.

The art wasn't exactly to my taste -- a little "picasso-esque" in a way that I personally don't appreciate. I'm not sure if there's a right way or a wrong way to draw an "athletic" build.

Even though I'd never use it to rank ladies, I could totally see how I might have (in my single days) used the same thing to choose amongst some boys that I used to be interested in. The amount of time spent nitpicking the framing story (I appreciated that they tied it to a historical figure and his actual writings) is honestly pretty weird to me.
posted by sparklemotion at 5:41 PM on June 22, 2014 [3 favorites]


I would be more willing to overlook the tonedeaf sexism in the article if there was some more meat to it. All I learned is that 36.8% is a magic number (admittedly a somewhat common problem with such methods - in the end it all boils down to checking aganst a critical value, and nobody bothers understanding any of the theory). If the framing story is just an excuse to talk about the general problem - and the cases where a probabilistic approach actually makes sense, nobody marries a large number of partners under repeatable conditions - the article could use some more actual discussion.
posted by Dr Dracator at 11:47 PM on June 22, 2014 [2 favorites]


My previous opinion of Krulwich, who is among other things a sexist of the first rank, is unchanged by this.
posted by spitbull at 7:37 PM on June 23, 2014 [2 favorites]


I agree with the above comments: the article is definitely problematic in several ways. But, from a purely mathematical point, the most interesting point here (which nobody has commented on yet!) is that the number "e" should make an appearance. (The article should have gotten a little more into the underlying math, I think.)

and, sparklemotion, it's an algorithm for maximizing, not satisficing. satisficing is arguably more interesting and relevant, but that's not what this analysis is about
posted by crazy_yeti at 12:55 PM on June 24, 2014


« Older The Trooper, an interpretation.   |   A hue angle of 270 degrees, a saturation of 50%... Newer »


This thread has been archived and is closed to new comments