UM Math on-line homework: Details

A more verbose analysis of the different points desired in a homework system, with some additional thoughts inspired by playing with the creation of some homework assignments. The following comments are rated with two numbers, M:N, where M gives the relative importance of the comment (2 = very, 1 = some, 0 = little) and N the degree to which the issue is better addressed by Webwork (2), either (0), or eGrade (2).

base characteristics

The automatic grading of any submitted homework
I like the ability to go and edit solutions in a homework problem after submitting them, as is allowed in Webwork, from the point of view that it allows the correction of minor typographical errors without having to log back in. When credit is reduced for each submission this is less significant, but may still be desirable. eGrade does not allow this, as the submission of the homework terminates the work session.
2 : 0
That each student get a homework problem set that is some way different from those given to other students
This is clunky or impossible to do in eGrade, and works well in Webwork.
2 : 2
That each student be able to rework errors in submitted homework, ideally for reduced credit
Webwork allows this. It's clunky to do with eGrade. On the other hand, it's probably sufficient to restrict the number of attempts as well; say with Webwork we allow six attempts, giving for successive ones 100%, 100%, 70%, 50%, 30%, and 10% credit. Then it may be reasonable if using eGrade to just restrict the number of attempts to three, which is easy to do, and leave it at that.
2 : 0
That it interface with some sort of on-line tutorial system that would allow students to do additional work on areas in which they are weak
Off the top of my head, it isn't obvious how to do this really well for either of the two systems. It's not essential for a homework system, but in the long-term the flexibility to do this would be nice. It's possible to write graders for problems in Webwork that could return a link to another "problem set," the tutorial, depending on the result on the problem, but this wouldn't be able to accomplish a similar thing based on the result for the entire set. eGrade grades the entire set at the same time, which could admit the possiblity of having a 'diagnostic grader' at the end, but it's not clear how this could be done automatically.
1 : 0
That instructors be able to view results for the homework easily, and download the results in tab-delimited or csv format for easy incorporation into a spreadsheet
Both systems do this well. It's annoying that it's not possible to have multiple sections in one class in eGrade (because that makes gathering aggragate data difficult), and annoying that data for individual sections in a class in Webwork don't appear to be separately downloadable.
2 : 0
That it include an easy way of generating problems for any given course
Overall it's easier to edit problems in eGrade than in Webwork.
1 : 2
That it support multipart questions
Both eGrade and Webwork do this, but eGrade imposes more formatting constraints on what the problems look like. This means that the look of the problems is more consistent from problem to problem, but sometimes makes it difficult to make the equations resemble the book problems that they are replacing.
1 : 1

other characteristics

That it maintain open/close dates for assignments, restricting access to times between those two
Webwork does this by default. I can't see how to do it in eGrade.
1 : 2
That it be easy to generate a formatted hardcopy version of the homework set
Webwork does this by default. The only way I can see how to do this in eGrade is by printing off the Web browser, which works but isn't as elegant.
1 : 2
(maybe) That it give students the ability to easily work on one or two problems at a time, without having to submit the remainder for grading simultaneously
Does this matter? If the problem set is regarded as something that the students work on their own, coming to the computer to enter their answers at irregular intervals, then this may be desirable--because it retains the state of each problem separately, so that when the student returns he or she doesn't have to re-fill in all of the answers and can do only those he or she is still working on. If they're just working on the problem set all at once and expect only to submit it once or twice, it doesn't matter as much.
0 : 2
That it be possible to include equations in fill-in-the-blanks questions
This is sort of piddling, but I ran into it in the set I was doing: "Write the equation of a circle with center (2,2) and radius 4: [_________] = 0." Being able to put the "=0" at the end allows the fill-in-the-blank to be a formula we can evaluate to test while having a more-or-less elegantly worded problem.
0 : 2
That multiple, equivalent, answers be allowed for a problem, or that custom answer evaluators can be added to the system.
This stems from another little thing that I ran into. I want to ask the question "Write an inequality for the half-plane below the x-axis." Answers include "y < 0" and "0 > y", which are testable as strings provided we can allow either as a solution. Webwork would, presumably, also allow writing a special evaluator to test this the inequality, because each question is tested for correctness by a Perl routine. To accomplish the same thing in eGrade would require a plug-in java module that talks the right API. This is theoretically feasible, but a different order of magnitude in difficulty.
1 : 2