* Zach White <[email protected]> [2007-06-14 15:05]: > On Thu, Jun 14, 2007 at 11:53:39AM +0200, A. Pagaltzis wrote: > > A surprising number of applicants can't actually program, so > > you want to weed them out early by asking for direct proof > > that they can do so. However, you also want to avoid the > > testing situation bias, so I wouldn't to sit them down with a > > moronic coding test (like that "FizzBuzz" silliness) during > > the interview, either. > > I find that people who speak poorly of FizzBuzz style tests > don't understand the problem they solve, or think they somehow > test for competency.
I can't speak for everyone who criticises FizzBuzz, but I didn't think (or claim) that that's the point of such a test. > FizzBuzz doesn't tell you how good a programmer is. It doesn't > tell you what their work ethic is like. It doesn't tell you how > well they work with others. It simply tells you that they can > solve a tiny programming problem, which an amazingly large > number of "programmers" simply can't do. Yes, and that is exactly what asking them to bring along code and then talk about it is also meant to accomplish -- no more than that. > It's not a general tool, it's only used to weed out those > people who come from untrustworthy sources, like a recruiter, > or craigslist. If a candidate is coming to me through one of > those sources it tells me that they don't have the ability to > get a job through the normal means (old coworkers, college > buddies, etc) and therefore deserve more scrutiny. FizzBuzz is > a cheap and easy way to filter those candidates at phone > interview time. Exactly; the point of asking for a snippet of code and some discussion of it is merely and solely to weed out the same categories of duds that FizzBuzz is also meant to solve. But having them provide proof by solving a cookie-cutter problem during the interview puts the candidate in a stressful closed-ended situation; whereas seeing some code they previously wrote and hearing them talk about it accomplishes exactly the same thing, while letting the candidate pick familiar and comfortable ground, making it a casual open-ended situation. Of course, a fringe benefit is that in contrdistinction to Fizz- Buzz, you get extra opportunities to figure out the candidate. But that's just a nice bonus and not the point. Perceptions are hugely important in the hiring game, so you want avoid anything that might introduce large random inputs or systemic biases. Therefore it is important to minimise the stressfulness and maximise the neutrality of the situation: neither the candidate's behaviour nor the interviewer's opinion should be inordinately affected by irrelevant minutiæ. I agree with the premises and the goal that lead to FizzBuzz- style tests. I simply disagree that FizzBuzz tests are the best way to achieve that goal. Regards, -- Aristotle Pagaltzis // <http://plasmasturm.org/>
