Radford shared with us this probability puzzle of his from 1999:
A couple you’ve just met invite you over to dinner, saying “come by around 5pm, and we can talk for a while before our three kids come home from school at 6pm”.
You arrive at the appointed time, and are invited into the house. Walking down the hall, your host points to three closed doors and says, “those are the kids’ bedrooms”. You stumble a bit when passing one of these doors, and accidently push the door open. There you see a dresser with a jewelry box, and a bed on which a dress has been laid out. “Ah”, you think to yourself, “I see that at least one of their three kids is a girl”.
Your hosts sit you down in the kitchen, and leave you there while they go off to get goodies from the stores in the basement. While they’re away, you notice a letter from the principal of the local school tacked up on the refrigerator. “Dear Parent”, it begins, “Each year at this time, I write to all parents, such as yourself, who have a boy or boys in the school, asking you to volunteer your time to help the boys’ hockey team…” “Umm”, you think, “I see that they have at least one boy as well”.
That, of course, leaves only two possibilities: Either they have two boys and one girl, or two girls and one boy. What are the probabilities of these two possibilities?
NOTE: This isn’t a trick puzzle. You should assume all things that it seems you’re meant to assume, and not assume things that you aren’t told to assume. If things can easily be imagined in either of two ways, you should assume that they are equally likely. For example, you may be able to imagine a reason that a family with two boys and a girl would be more likely to have invited you to dinner than one with two girls and a boy. If so, this would affect the probabilities of the two possibilities. But if your imagination is that good, you can probably imagine the opposite as well. You should assume that any such extra information not mentioned in the story is not available.
As a commenter pointed out, there’s something weird about how the puzzle is written, not just the charmingly retro sex roles but also various irrelevant details such as the time of the dinner. (Although I can see why Radford wrote it that way, as it was a way to reveal the number of kids in a natural context.)
The solution at first seems pretty obvious: As Radford says, the two possibilities are:
(a) 2 boys and 1 girl, or
(b) 1 boy and 2 girls.
If it’s possibility (a), the probability of the random bedroom being a girl’s is 1/3, and the probability of getting that note (“I write to all parents . . . who have a boy or boys at the school”) is 1, so the probability of the data is 1/3.
If it’s possibility (b), the probability of the random bedroom being a girl’s is 2/3, and the probability of getting the school note is still 1, so the probability of the data is 2/3.
The likelihood ratio is thus 2:1 in favor of possibility (b).
Case closed . . . but is it?
Two complications arise. First, as commenter J. Cross pointed out, if the kids go to multiple schools, it’s not clear what is the probability of getting that note, but a first guess would be that the probability of you seeing such a note on the fridge is proportional to the number of boys in the family. Actually, even if there’s only one school the kids go to, it might be more likely to see the note prominently on the fridge if there are 2 boys: presumably, the probability that at least one boy is interested in hockey is an higher if there are two boys than if there’s only one.
The other complication is the prior odds. Pr(boy birth) is about .512, so the prior odds, which are .512/.488 in favor of the 2 boys and 1 girl, rather than 2 girls and 1 boy.
This is just to demonstrate that, as Feynman could’ve said in one of his mellower moments, God is in every leaf of every tree: Just about every problem is worth looking at carefully. It’s the fractal nature of reality.
I first encountered this problem in David MacKay’s “Information Theory”. He doesn’t mention that it’s originally from Neal. (Neal is cited extensively elsewhere in the book for more substantial contributions, and thanked in the acknowledgments.) Is this standard practice with problems? When I’ve given this problem to students, I’ve always said that it came from MacKay’s book, and recommended that they look in it for more problems like it. If I had known of the original source, I could have recommended that they check that out too, and I also would have known to look at it myself.
I think there is a problem with the solution this brain-teaser, that you touch upon gentle-heartedly but it actually a real problem.
When you are calculating P(Evidence 1 & Evidence 2 | xGyB), where Evidence 1 is seeing a girl’s bedroom, and Evidence 2 is seeing a boy’s letter (and xGyB means x girls and y boys), we can actually break down each Evidence even into 2 events; namely, the event that the evidence *exists*, and the event that we see that evidence. So for Evidence 1, we can think about P(The neighbors have a room set up for a girl & I stumble into that room). With the quaint old fashioned notions of pink girl’s room, the first half has probability 1 (whenever they have a girl) and the second half has the same probability as the fraction of girls.
But for Evidence 2, it’s a little more difficult. If it breaks down into P(the neighbors get the boys letter & I see the letter), the first half does evaluate to 1 (under pretty reasonable assumptions), but why do we assume the second part evaluates to 1 as well? It seems like it is a probability that can’t be handled with the theoretical niceness of “P(1G) = .5”.
I guess you could add more conditionals, like saying that P(I stumble into that room) is actually P(I stumble into that room | I stumble into one of the bedrooms). Then we could say that P(I see the letter) is actually P(I see the letter | I see a gender-related child letter). But then Evidence 1 still simplifies to merely a question of what fraction of the children are girls, but Evidence 2 seems totally indeterminate, because we can’t assign simple probabilities to things like “looking at a refrigerator” or “the boy’s letter is on top”
So is the simple answer (1/3 vs 2/3) correct? Or is your point that because you can’t reduce the second part of Evidence 2 to a ratio of # of successful states / # of total states, that the problem cannot be solved stylistically (like the old “Tuesday’s boy” problem), but rather you have to deal with the messiness of reality?
it’s a nice problem I guess with the intention that one piece of evidence is drawn at random, and the other piece of evidence is fixed. Perhaps it would have been cleaner to just have the neighbors confirm they have at least one son? Then it’s a direct analogue to the monty hall problem, complete with the three bedroom doors instead of three game show doors.
But, if the kids went to three different schools, and there were three similar letters on the fridge (where the other letters could be about the girls ice hockey – er, that is field hockey – team… and then what are the chances that a boy letter was on top? Then we’d have a three-letter problem in addition to a three-door problem.
Maybe the fact that all 3 children come home at the same time is a clue that they all go to the same school…
BTW, getting down the “every leaf” road, using sex ratio at birth as a base rate for 2+1 vs. 1+2 assumes that people are not conditioning their procreation on the sexes of the already born children. Statistics on family structure must be enormous and relatively easy (not for me) to find, so that supposition is quite testable.
I’d assume that there is a non-negligible probability that the school principal sent such a letter to the wrong parent, so as far as I’m concerned, 3 girls should still score nonzero. Probably one could also come up with a possible explanation for 3 boys as well, although I’d give that less probability.
OK, I probably violate “You should assume all things that it seems you’re meant to assume” here, but then it’s just more leafs for god to hide in.
Njnnja: I think the complications you mention don’t affect the answer. As long as the probability of the letter being put on the refrigerator does not depend on whether they have one boy or two, it cancels out in the odds ratio. (And if you don’t assume that, it’s entirely unclear whether the probability would be higher or lower with two boys rather than one, so it’s covered by the last paragraph.) And it’s not actually necessary to assume that any girl’s bedroom would have the traditional appearence seen. If you assume that there is some probability of a girl setting up her room that way, you will find that this probability also cancels in the odds ratio. There maybe could be a problem if when there are two girls, whether one girl sets up her room that way is not independent of whether the other girl does.
By the way, I take “God is in every leaf of every tree” not so much as a saying that there is infinite complexity in every situation, but as saying that every aspect of this complexity is interesting, in a fractal sort of way. In the context of career advice, the answer to a question such as “Should I specialize in statistical genetics, or work on a variety of applications” is “either”. If you spend your career on just statistical genetics, there will be plenty to work on, and the details that you see that you only see from having an in-depth knowledge will be interesting. But if you instead work on many applications, that too will be interesting, and you wil see interesting connections that someone working in only one area would miss…
I am confused about this:
“(a) 2 boys and 1 girl, or
(b) 1 boy and 2 girls.
If it’s possibility (a), the probability of the random bedroom being a girl’s is 1/3, and the probability of getting that note (“I write to all parents . . . who have a boy or boys at the school”) is 1, so the probability of the data is 1/3.
If it’s possibility (b), the probability of the random bedroom being a girl’s is 2/3, and the probability of getting the school note is still 1, so the probability of the data is 2/3.
The likelihood ratio is thus 2:1 in favor of possibility (b).
”
For example, for (a), I get that the probability of the random bedroom being a girl’s is 1/3, but I don’t get why the probability of getting the note is 1. Would that mean that every family that has a boy has the same note on the fridge? I have 3 boys and there is no such note on my fridge. What am I missing?
The probability of the note being pinned to the fridge is not the issue. The note on the fridge is intended as a device to inform you that the family has at least one son. The problem could have been written to exclude the note, instead having one of the hosts mention that their son, David, wouldn’t be joining because he was away at hockey practice. Either way, you are being told that there is a son, but in a manner that doesn’t give you information about the existence of a second one. This contrasts with the way in which you learn about a daughter, which is a variation of the Monty Hall problem, as Eric notes.
Of course, in keeping with the point of the post, the phrasing and inflection of the bit about David could tilt the scale one way or another. A parent might emphasize the word “David” if David were one of multiple sons, and “son” if not. Eliminating the note may have pruned one leaf but added a new one. This is why a physicist would simply declare David to be a frictionless sphere and move on.
No, mentioning that their son David wouldn’t be there due to hockey practice would not be at all similar. If they had two sons, they might very well both be at hockey practice, and the host would presumably have said that both would be absent. And on the other hand, not all boys play hockey, so the probability that a son will be missing due to hockey practice is higher with two than with one. Plus, if hockey practice conflicts with dinner time, it’s possible that one or more girls could be absent due to hockey practice, so the lack of any statement about an absent daughter is informative. So the probability of the observed utterance depends very much on whether they have one son or two, unlike the refrigerator note (making some reasonable assumptions).