A lot of these posts don't actually address the problem of hiring and filtering people. If you think whiteboard tests or logic puzzles are not effective, why do you think this and what is a better alternative?
I personally think they are ok, but should more be a means to test if the candidate can reason intelligently and it shouldn't matter much if the end answer they get is correct or not.
Personally I think there are two good ways, in combination, to determining if a person is a good hire.
1. Ask questions about past projects. Probing questions to make sure their resume sounds accurate. Get a sense for how they solved problems in the past and explain them.
2. Give them something to code. On a computer. Some small example. Even better if you can try a pair programming exercise to see how they code and how they can communicate with others on a team.
Asking academic questions has been entirely useless in my opinion. I've worked with multiple people who can pass all the academic questions about algorithms and data set and couldn't write an application to save their live...and the vice versa too. I'd rather see how they can code, now and how they can communicate about projects (active and in the past).
I always try to find out if the candidate could - with enough information - solve the puzzle.
Can you, given time, figure out a good solution to the problem I am asking? How would you go about solving the problem? What is your methodology?
I agree with you. These sorts of problems are really to understand how someone reacts to a new idea that he or she does not understand. A lot of times people tell me that they don't know they answer to a problem, which is fine. Knowing that there are things that we don't know if a sign of maturity, and the curiosity to figure out how to solve a problem is what we should be looking for.
I personally think they are ok, but should more be a means to test if the candidate can reason intelligently and it shouldn't matter much if the end answer they get is correct or not.