Re: [WWdevel] Re: extraAnswerEvaluators.pl

[Davide P.C. <dp...@un...>]

> Just one additional comment about this behavior.  I agree that in 
> cases where an answer is of smaller magnitude than the absolute 
> numerical tolerance, it is mathematically reasonable to treat the 
> answer as zero. I am a little worried that students will think that 
> the fact that they entered an answer that is non-zero and it is 
> previewed as zero will make them think that the system isn't working 
> correctly, however.  For example, if the correct answer is evaluated 
> with num_cmp(sqrt(7)) and the student enters 1/sqrt(7), the system 
> reports
>
>     Entered	Answer Preview	Result
>     0		 1/sqrt(7)	incorrect

That would only be the case with num_cmp(sqrt(7),tol=>.5), or some 
tolerance bigger than 1/sqrt(7), not num_cmp(sqrt(7)), which would be 
using relative tolerances, so the zeroLevel tolerances would be used 
for determining zero.  Remember that this is ONLY a problem when 
ABSOLUTE tolerances are used, and when they are larger than the 
student's answer.

I'm willing to say that when absolute tolerances are used, don't change 
to zero, but it is important to remember that, as far as WW is 
concerned, if tol=>.5 is in effect, 1/sqrt(7) IS zero.  If you don't 
want 1/sqrt(7) to be considered as zero, the tolerances are set wrong, 
and if it is considered to be zero, why SHOULDN'T the student be made 
aware of that?

Personally, I think that displaying ALL answers using the precision of 
the problem would be a help, as then students would start to become 
aware of such issues, which they aren't, but I'm sure some of you will 
disagree.  :-)

Davide