The best interview question I ever got asked was, "What's your favorite algorithm?" I was completely and utterly taken aback: I'd never been asked such a thing before, and had no answer whatsoever.
So I reached back into my past hobby work, and pulled out "lottery scheduling". A 30-40 minute conversation on the description, performance, trade-offs, implementation trade-offs, features and misfeatures of lottery scheduling ensued.
This worked more brilliantly as a way to interview than anything else I'd ever been through, because it gave me and my experience a place to shine. Most interviews really don't. They are idiot-test after idiot-test, designed to wash out bad hackers, but in the process producing a crop of Most Assuredly Not Idiots who may not actually be good, while accidentally washing out some really good hackers who, for example (first-hand experience), wrote a recursive function instead of a while-loop and a Stack<T>.
If you want good, give good a chance to shine. If you want not-bad, keep stacking on the idiot tests.
I've asked, "what's your favorite part of math?" for the same reason. Something open-ended, but allowing the person being interviewed to be on their home turf. Seems more fair than only focusing on a topic I happen to know backwards and forwards, but that they, through happenstance, might not.
This can get at a part of creativity that's independent from questions about, say, the last project someone worked on.
One drawback with that phrasing, which is the one I've used, is that "favorite" can cause lockup, because choosing that one thing can be hard in real-time.
So I reached back into my past hobby work, and pulled out "lottery scheduling". A 30-40 minute conversation on the description, performance, trade-offs, implementation trade-offs, features and misfeatures of lottery scheduling ensued.
This worked more brilliantly as a way to interview than anything else I'd ever been through, because it gave me and my experience a place to shine. Most interviews really don't. They are idiot-test after idiot-test, designed to wash out bad hackers, but in the process producing a crop of Most Assuredly Not Idiots who may not actually be good, while accidentally washing out some really good hackers who, for example (first-hand experience), wrote a recursive function instead of a while-loop and a Stack<T>.
If you want good, give good a chance to shine. If you want not-bad, keep stacking on the idiot tests.