So people are complaining that whiteboarding is the wrong way to interview. That we don't discover if they can write and maintain complex software system, inste

So people are complaining that whiteboarding is the wrong way to interview. That we don't discover if they can write and maintain complex software system, instead focusing on 'trivia'. I don't disagree (I think that the trivia being tested can be important - a team with no one that has deep knowledge of data structures of algorithms is lost), but I also don't know a good way to interview that finds out what we need to know. How do you find out if someone can make good long term decisions in a short amount of time? It seems inherently contradictory. Some people say by giving homework problems. They're not wrong, that might give valuable insight, but I know that I would hate that, so others might too. Can there be a good answer?