[LON-CAPA-users] duplicate correct answer in randomized test

[Gerd Kortemeyer]

Hi Nathan,

On Mar 2, 2006, at 12:24 PM, Nathan Schoenack wrote:

> Right now I only have evidence of this happening once, but  
> yesterday we gave a randomized test where one test had a question  
> that had two identical, correct answers.  Has anyone else seen this  
> kind of thing with the randomized tests?  I'd like to find out if  
> this happened due to the way I coded the problem, or if this was  
> just a random glitch.

Yep, that happens frequently, and currently LON-CAPA has no safe- 
guard against it.

Ways how this can happen:

a) explicitly giving an array with wrong answers, but one of them is  
correct. This can happen when generating incorrect answers  
algorithmically, but in certain scenarios dividing the correct answer  
by one or things like that. Or if your correct answer is "c*a/b," you  
make "c*b/a" one of wrong choices, but "a" happens to be equal to "b"  
in certain situations.

b) setting the display format such that a randomly generated wrong  
answer is displayed the same as the correct answer. 3.104 and 3.102  
are displayed the same if you pick "3s" as your output format.

Either way, it's not a big deal.

a) works fine
b) works fine, too, except if the display format is accidently set  
such that the computer would find its own answer wrong because your  
tolerance is too low - but then it's a problem no matter what. You  
would get warnings in the editor, particularly if you had it  
calculate a bunch of answers for you


BTW, this can go either way. If I make two times the same wrong  
answer (at least in display), I get that twice. So just because an  
answer appears twice, you don't know yet if it is right or wrong.

The saving grace is that LON-CAPA grades in such way that it does not  
check on answers A, B, C, etc. Instead, it plugs the value next to  
the bubble into the normal numerical response, as if the student had  
typed in that answer. So if two or more answers are the same, they  
would lead to the same grading.

Once or twice a semester, I goof up in this way on a question. When  
students freak out during the exam, I tell them that they are lucky,  
because they have one less choice, so their chance of randomly  
picking the correct answer is increased.

- Gerd.