[PRL] Computer science `education'

[Joe Marshall]

On Nov 9, 2007 1:17 PM, Peter Dillinger <pcd at ccs.neu.edu> wrote:
> On Fri, Nov 09, 2007 at 12:00:47PM -0800, Joe Marshall wrote:
> > Was I expecting too much?  We're all my questions way too hard?  Or did the
> > college completely fail in its mission?  Or is the general computer
> > science curriculum
> > this bad everywhere?
>
> are all your interviews this trite?

No, some are more trite, some are less.  It depends a lot on what I can glean
from the resume and other pre-interview material.  In the case of this
candidate,
all I had was basically a list of classwork he did.  I also asked a few other
questions that I didn't mention earlier.

> sure, you want to know whether the candidate has good vocabulary and
> working knowledge of his field, but you can find this out without
> making the interview into a quiz.  and there's much more you want to
> find out!

I know!  I was hoping that one of these questions would trigger a conversation.

> rather than asking a "quiz" question about advatanges of functional
> programming,  how about giving him several scenarios (e.g. static
> analysis tool; one time data conversion; physics simulation) and ask
> what language he would prefer to use and why.

That's an interesting idea, I'll think about that.

> how about disassembling something:  ask him to describe in as much
> detail as he wants, focusing on the parts he finds most interesting,
> the chain of events that happen from pressing Enter in a web browser
> (after typing a URL into the address bar) and a web page appearing.

That's a fairly common question around here, so I couldn't use that one
directly, but I could probably come up with a similar variation.

> i guess one thing i'm promoting here is to avoid questions that could
> lead to an "i don't know" response.  let the person talk about what
> they do know.

Yes, but that is why I had a series of questions.  I wanted at least one
to elicit a response.  After I went through these, we did talk about brittlestar
neuronal photosensitivity (his research work).  It was cool, but not really
relevant.

> also, if the point is to find out how well they perform a certain job,
> i think that presenting the candidate with scenarios he might
> encounter in that job is the most valid way to predict what his
> success/productivity would be.  in my internships, i've never had
> anyone ask me what alpha-beta pruning is.  everyone i've worked with
> was smart enough to look up such things themselves.
>
> on the other hand, i have had people ask me if i could help find the
> bug in their code.  i have had people ask for my expert opinion on
> how to do something.  i have had people ask for my critique of their
> work.
>
> and time doesn't permit asking someone to solve the kinds of multi-week,
> multi-month, or multi-year problems he/she will be working on, but you
> can ask about problems they've solved before.

The problem is that during these `batch' days, I only get 1/2 hour with
the candidate and there isn't much you can do in that amount of time.

> my $0.01859 Canadian.
>
> with an addendum!
>
> one could argue that if someone is dedicating brain cells to remembering
> stuff he or she is not currently dealing with, he's not meeting his full
> potential in terms of reasoning power, creativity, or general
> problem-solving ability.  there is some science behind this "finite
> capacity" theory, and on the anecdotal front, i find that professors
> (my token "smart" and "capable" population for this point) are quick
> to purge things they don't need from their memory of "well understood"
> things and skilled at doing so.

Yeah.  I wish I could use the brain cells that remember the lyrics and
tune of the Gilligan's Island theme (both versions) for something a bit
more useful.

> and one final disclosure:  i've never taken or been offered a job i had
> to interview for.  :)

Do you want to?  Google is hiring.....


-- 
~jrm