The only interview questions I've ever found at all reliable--in the sense of providing good answers to your questions about comparison & assessment--are coding problems. As in, here's a toy problem, you have 45 minutes/3 hours to write some code to solve it, on this actual computer with actual Google etc.
That actually tests a set of skills, which puts it well ahead of various interview questions which test nothing of relevance. But it's also a limited set of skills, some of which a recent graduate might reasonably lack but might be reasonably be expected to obtain quickly (if the right kind of smart, committed, a quick study etc).
The problem is I don't know really know how to test for anything but current ability to write reasonable code for a not-too-large problem. The more experience I get interviewing, the more decisions feel arbitrary to me. And I'm pretty skeptical when other people claim to have it figured out.
That actually tests a set of skills, which puts it well ahead of various interview questions which test nothing of relevance. But it's also a limited set of skills, some of which a recent graduate might reasonably lack but might be reasonably be expected to obtain quickly (if the right kind of smart, committed, a quick study etc).
The problem is I don't know really know how to test for anything but current ability to write reasonable code for a not-too-large problem. The more experience I get interviewing, the more decisions feel arbitrary to me. And I'm pretty skeptical when other people claim to have it figured out.