> A false positive in the interviewing process (hiring a dud) is much more likely than a false negative (turning off a great candidate).

I see statements like this thrown around a lot. Do you have precision/recall/F-measure numbers to back this up?

The most famous example I know of is the guy who implemented a FizzBuzz test and thereby eliminated the vast majority of programmers who otherwise appeared qualified on paper and in person. I don't have the link handy but I think it was covered in codinghorror.com.

I have been very interested in running a similar experiment myself, but alas I am not a hiring manager. As I recall from that example, there were no details of what counted as "qualified" on paper and in person.