[CommieBobDole]

My thought is the purpose of a job interview is to weed out people who are an obvious poor fit for the job temperamentally, and to verify that the interviewee has a working understanding of the relevant skills that they listed in their resume. Beyond that, actual job performance depends on a lot of factors that are very difficult to assess in an interview.

I think a lot of companies have deluded themselves into thinking they've designed some sort of unique interviewing system that can guarantee good job performance (instead of reducing the incidence of catastrophically bad performance), but honestly, job performance itself is very hard to measure in any objective way once you disregard the extremes.

Just because a misused tool is bad at doing the thing you're misusing it for doesn't mean that it doesn't have utility when used for its actual purpose.

[ipnon]

You can just get on the phone with them. Talk shop for an hour. Tell them about problems your company is having. Ask them how they would solve them. An experienced engineer can determine a person's relative skill level this way.

Algorithm quizzes aren't the only way to reduce false positives. Can we say definitively that they are the best?

[CommieBobDole]

I agree with you - the best way to do the things I talked about is to spend time talking with the person to ensure that they actually understand and can reasonably communicate about the skills they claim to have.

There's probably a place for practical programming tests, too, but whiteboarding algorithms and such seems to be more of a fraternity hazing/secret handshake ritual than anything else. Good if the skill you need to hire for is "Read and memorized 'Cracking the Coding Interview'", not really indicative of anything else.