Comments on: What software is not http://williamtozier.com/slurry/2005/10/26/what-software-is-not Pontification without all the gritty gravitas Sun, 20 Jul 2008 00:05:19 +0000 http://wordpress.org/?v=2.5.1 By: Alex http://williamtozier.com/slurry/2005/10/26/what-software-is-not#comment-156 Alex Wed, 26 Oct 2005 20:58:20 +0000 http://williamtozier.com/slurry/?p=156#comment-156 With respect to your initial point about firing HR people who use these sorts of tests: the norm in software industry interviews is to ask people to write code, on the whiteboard, so you'd be firing a lot of people =) That said, what the interviewer [if s/he is good] will be looking for, apart from the code itself, is what the interviewee leaves out or assumes: error checks, boundary cases, string lengths, Unicode vrs ASCII etc. In other words, all the things that matter when you're writing software for the real world. Somebody who doesn't display any awareness of these factors tends not to get hired. It's not an ideal assessment tool, but in my experience it tends to do a reasonable job, possibly at the expense of a high false negative rate ie ppl rejected who could have done well. From that perspective, it might not be entirely unreasonable to ask people to write code during an exam, but only if they are actually also graded on how they deal with these ambiguities inherent in the fact that software has to interact with the rest of the world [in contrast to math problems which tend not to have ambiguities, unless placed there by a sneaky professor ;-)]. With respect to your initial point about firing HR people who use these sorts of tests: the norm in software industry interviews is to ask people to write code, on the whiteboard, so you’d be firing a lot of people =) That said, what the interviewer [if s/he is good] will be looking for, apart from the code itself, is what the interviewee leaves out or assumes: error checks, boundary cases, string lengths, Unicode vrs ASCII etc. In other words, all the things that matter when you’re writing software for the real world. Somebody who doesn’t display any awareness of these factors tends not to get hired. It’s not an ideal assessment tool, but in my experience it tends to do a reasonable job, possibly at the expense of a high false negative rate ie ppl rejected who could have done well.

From that perspective, it might not be entirely unreasonable to ask people to write code during an exam, but only if they are actually also graded on how they deal with these ambiguities inherent in the fact that software has to interact with the rest of the world [in contrast to math problems which tend not to have ambiguities, unless placed there by a sneaky professor ;-)].

]]>