Saturday, December 13, 2008

On Grading Algorithms

I finished grading the finals (graduate algorithms) before I left on vacation. Grading algorithms finals has its
  • challenges: grading is not just checking "equality". A student might propose a solution different from the one you have in mind but equally valid. So you have to understand their idea, think about if it will work, check proof of correctness and make sure there are no bugs, etc. All time consuming tasks.
  • surprises: a silent student in the classroom may show a spark in the finals. A regular class room participant may fumble solutions to easy problems.
  • and, in this case, self-inflicted quirks: borrowing from someone (who? I don't now recall), I award 1/4th of the points on a question if a student simply writes "I Do Not Know".This rewards the students for knowing they do not know the solution, as well as cutting down on fluff a grader has to read through. How do students react to this? Some are professional, fill in "I Do Not Know" just before the time runs out, some are principled and never use this route, and yet others forget to use this route and regret it later.

Labels:

12 Comments:

Anonymous Anonymous said...

Muthu,

I also use the policy of 1/4 credit for the answer "I don't know". I like the policy well enough to keep it, but it is not perfect. I find that most students are not so good at estimating the probability that they can give a correct answer.

Cheers,
Kirk Pruhs

1:37 PM  
Anonymous Anonymous said...

I first heard the 1/4 policy (for some value of 1/4) from Martin F-C. Not sure if he was the original.

1:43 PM  
Blogger JeffE said...

I've been giving 25% for "I don't know" (or "no idea", or "WTF", but definitely not " ") for about ten years. I don't remember who gave me the idea.

Most students seem to appreciate the WTF policy. (Yay, free points!) But a few (always undergrads) think it's just a sadistic mind game, because they never know whether their partial answers are worth more or less than "WTF". Encouraging those students to spend more time thinking about solving the problem instead of strategizing about partial credit just pisses them off.

6:55 PM  
Blogger Rogatienne said...

Ciao Dude,
it might be the case that you heard the the policy of 1/4 credit for the answer "I don't know" from Allan Borodin in Rome?
It is my case, and maybe we talked about it...

Ciao!

2:58 AM  
Blogger Massimo said...

I heard about it from Allan Borodin in UofT.

6:17 AM  
Blogger Unknown said...

Hi, Well probably someone else used
this type of scheme before me, but I have been using it for many many years at Toronto. (I probably told Muthu when we were both enjoying great academic visits in Rome, La Sapienza.) I give 20% for any question for which a student writes "I do not know how to answer the question". I also allow students to say "I do not know how to do this problem set/text/exam" and get 20% for the entire set/test/exam. In recent years I have been giving 10% for those who write nothing. This goes against the pedagogy of "giving credit for knowing what you dont know" but it also cuts down on those who get flustered at the end (esp in a test), forget to write it, and then beg. I have found that students love this system as they feel you are being generous and good students recognize the value in not bullshitting. I rarely get nonsense since using this 20% rule.

9:52 AM  
Anonymous Anonymous said...

I first heard about it from Naomi Nishimura in 1991 when co-teaching a course with her at Waterloo. We wanted to ask "essay" like questions but we were worried about lengthy responses from the students. At some point the concept of giving students 20% for a blank question came up.

I don't know if she got this idea from from her grad years at U of T where Allan is or if the flow of information was in the opposite direction.

Alex Lopez-Ortiz

2:56 PM  
Anonymous Anonymous said...

I now distinctly remember Allan telling me about the nuances of this scheme, as Rome, the eternal city, blinked, in the evening. Thank you, La Sapienza!

The scheme -- its variants upto O(1) factors --- cuts down bullshit, discourages strategizing for partial credits, but seems to need the students to get used to the idea some, and estimate the quality of their solutions a little better.

I also liked the tweak of awarding 20% for no answers: to avoid students kicking themselves for not writing IDNTK. Have to see how that works, next semester.

-- metoo.

4:24 PM  
Anonymous Anonymous said...

In my grad classes I combine this with the policy of asking the students to answer any X of the Y questions on the test. In general I would rather that students have a full understanding of 1/2 of the material than have a 1/2-***** understanding of all of the material.

Cheers,
Kirk Pruhs

9:31 AM  
Anonymous Anonymous said...

To followup on my previous message, we found that it was important to warn the students about this marking scheme well ahead of time. We allowed them to ask questions in class about how would this work. Overall it was very effective in stopping students from fishing for marks.

Alex Lopez-Ortiz

p.s. I asked Naomi and she can't remember if this is something that occurred to her on the spot or if she had heard about it from someone else.

7:21 PM  
Anonymous Anonymous said...

My last word on the subject. Prabhakar Ragde who did a postdoc in Toronto remembers hearing this idea as far back as 1987 from Charlie Rackoff, who might well have been quoting Al.

In sum, all sources lead to Rome, i.e. Al Borodin.

9:11 AM  
Anonymous Anonymous said...

Muthu, is this new for your class? And specific to algorithms? I took your databases course once, and no such policy! Or is it that databases doesn't deserve such grace :-)

9:15 PM  

Post a Comment

<< Home