When I applied for my first job they had one of those quasi Microsoft questions. It's a scenario based on real life to see if you know how to make good decisions, and it goes like this:
You're making films for some leading company X. Recently people have complained your films have been corrupted and not producing the right results and this is bad. So you did some research and you found out your factories have poor standards and your workers are all hired last week who have no idea what they're doing. One of the factory had an explosion recently. Your scientists analyzed the corrupted samples and says lab result shows you can only get that with radiation, and for whatever reason your factories do have radioactive stuff sitting around in the one that just blew up recently.
Given this, what would you do?
So I said something like close off the factory and send in someone to test for radiation until the cause can be isolated, and that turned out to be totally wrong. The correct answer was that the scientists were wrong and it's just your stupid workers sucking, and in fact if you answered anything related to radiation you got docked major points since it takes a lot of $ to send in the specialists to deal with them. You're supposed to ignore that because based on years of experience in field it is extremely unlikely a factory would have a radioactive source strong enough to mess up all your film.
Now that's all good as a case analysis if I really was an expert in the film-making industry, but I am NOT a guy who has been in the film industry for years. Now most 'Microsoft' questions aren't as bad as this, but still after reading about them I think most of them are just kind of a 'gotcha' thing, and getting them right is more about researching enough on the Internet ahead of time to see all the answers as opposed to actually knowing something. Worse yet, sometimes it'd appear the correct answer is just the one whoever has thought of. Here's another question from the same company that they use:
You have 7 black jellybeans and 7 red jellybeans. You have to distribute them in two piles in any order, but they all have to be used. You then grab a jellybean from each pile at random. What is the distribution that maximizes the chance of getting 2 black jellybeans?
Now if you distribute them evenly then it's 1/2 * 1/2 = 25%. Now anybody who has done a question should suspect there's a catch and look deeper, and you'll realize that you can just do 1 black jellybean in one pile and rest in the other pile, and then it's 1 * (6/13) = 6/13 = ~50%. That was what I answered and it was the correct answer.
But wait, how do you know it is the maximum possible? You DO NOT! When I answered the guy was just like 'yep that's correct', but after I went home and thought about it, he didn't ask me to prove you cannot do better than that. To do so you'd have to take the express the probabilty of getting two black beans in two variables (number of black on side 1 and number of red on side 1) and then take the derivative of that. And I think if you do all that you indeed find that this solution is optimal. But that's actually really tough and certainly it'd take way too long to come up with the complete solution in the time you have. So basically all this question is testing is did you realize you could put just 1 black jellybean on one side? Yet there's really no guaranteed doing that maximizes the chances of getting two blacks, at least it's certainly not obvious unless you have did some rigorous modeling of the problem.
I find that it's easier to just google 'Interview questions' and memorize all the answers than actually trying to solve them. Now obviously if the guy wants you to write a program that does sorts roman numerals in Java it's not something you can memorize ahead of time, but there does appear to be a relatively limited bank for the 'Microsoft' questions. So whether you get some job may be dependent significantly on whether you memorized a bank of questions? That seems pretty messed up to me.
You're making films for some leading company X. Recently people have complained your films have been corrupted and not producing the right results and this is bad. So you did some research and you found out your factories have poor standards and your workers are all hired last week who have no idea what they're doing. One of the factory had an explosion recently. Your scientists analyzed the corrupted samples and says lab result shows you can only get that with radiation, and for whatever reason your factories do have radioactive stuff sitting around in the one that just blew up recently.
Given this, what would you do?
So I said something like close off the factory and send in someone to test for radiation until the cause can be isolated, and that turned out to be totally wrong. The correct answer was that the scientists were wrong and it's just your stupid workers sucking, and in fact if you answered anything related to radiation you got docked major points since it takes a lot of $ to send in the specialists to deal with them. You're supposed to ignore that because based on years of experience in field it is extremely unlikely a factory would have a radioactive source strong enough to mess up all your film.
Now that's all good as a case analysis if I really was an expert in the film-making industry, but I am NOT a guy who has been in the film industry for years. Now most 'Microsoft' questions aren't as bad as this, but still after reading about them I think most of them are just kind of a 'gotcha' thing, and getting them right is more about researching enough on the Internet ahead of time to see all the answers as opposed to actually knowing something. Worse yet, sometimes it'd appear the correct answer is just the one whoever has thought of. Here's another question from the same company that they use:
You have 7 black jellybeans and 7 red jellybeans. You have to distribute them in two piles in any order, but they all have to be used. You then grab a jellybean from each pile at random. What is the distribution that maximizes the chance of getting 2 black jellybeans?
Now if you distribute them evenly then it's 1/2 * 1/2 = 25%. Now anybody who has done a question should suspect there's a catch and look deeper, and you'll realize that you can just do 1 black jellybean in one pile and rest in the other pile, and then it's 1 * (6/13) = 6/13 = ~50%. That was what I answered and it was the correct answer.
But wait, how do you know it is the maximum possible? You DO NOT! When I answered the guy was just like 'yep that's correct', but after I went home and thought about it, he didn't ask me to prove you cannot do better than that. To do so you'd have to take the express the probabilty of getting two black beans in two variables (number of black on side 1 and number of red on side 1) and then take the derivative of that. And I think if you do all that you indeed find that this solution is optimal. But that's actually really tough and certainly it'd take way too long to come up with the complete solution in the time you have. So basically all this question is testing is did you realize you could put just 1 black jellybean on one side? Yet there's really no guaranteed doing that maximizes the chances of getting two blacks, at least it's certainly not obvious unless you have did some rigorous modeling of the problem.
I find that it's easier to just google 'Interview questions' and memorize all the answers than actually trying to solve them. Now obviously if the guy wants you to write a program that does sorts roman numerals in Java it's not something you can memorize ahead of time, but there does appear to be a relatively limited bank for the 'Microsoft' questions. So whether you get some job may be dependent significantly on whether you memorized a bank of questions? That seems pretty messed up to me.
