All entries for Monday 23 January 2006
January 23, 2006
I have been exploring two very different assessment programs recently and was drawn to compare the way they each handle Multiple-Response Questions (MRQs). To put things in context, consider the following naive example of such a question:
Decide whether the following arithmetical statements are true or false:
(For non-arithmeticians, this fourth part is the only true statement.)
The two packages impose their own different marking schemes and I am happy with neither. Here are their inflexible offerings:
This software is a simple quiz builder, very easy to learn and quick to author. (If you have the questions ready, you could put together a 10-question quiz in 15 minutes, even first time round.)
You answer an MRQ like the one above by checking all the buttons of the statements you think are true and leaving unchecked those you think are false — the buttons toggle on and off like conventional check boxes. Full marks are given if and only if every part is answered corrrectly (with true statements checked and false ones unchecked); otherwise zero is given.
I feel that this all-or-nothing approach is too severe; a student gettting three parts out of four right surely deserves some reward.
In contrast to the previous package, this one is a behemoth, powerful but hard to tame. (Incidentally, I notice that, unlike the "alleluias" and "slaves" that eBay claimed on Google to be auctioning before Christmas, today it doesn't appear to have any behemoths for sale.) Package B's multiple response offering is part of its MCQ environment — you move from MC to MR by simply ticking the box "allow more than one correct answer". Below each MRQ, a hyperlink partial grading explained appears in red; when clicked, the following message pops up in a new window:
What does formula mean? Are "correct choices" the same as "correct answers"? If not, then perhaps "# correct answers" means "# of true statements"? And what if the grade is negative? It's not clear. (Incidentally, it would save us all a lot of time if questions really "could calculate their grades".)
Imagine for simplicity that
- the above MRQ is the sole question on a test
- a desperate student in a rush launches the test and immediately presses the "submit" (or "grade") button, neither reading the question nor checking any buttons.
If we assume all the buttons start unchecked by default, the desperado gets 3 parts correct and one part wrong, and by the most likely interpretation of the formula, scores two-thirds out of a maximum one point; in other words, 66% !! That's certainly 'owt for nowt' and a fat reward for opportunism — hardly a desirable outcome.
What I would like
In neither package is the author given any choice about the format or a marking scheme for a multiple-response question. You must take it or leave it. But just in case the developers are listening, here is the kind of flexibility I would like to see as standard for MRQs:
- A drop-down menu in a combo box next to each part of the question with the three options: 'true', 'false', 'no attempt', and with all the boxes initially set to 'no attempt'.
- The ability to set the marks awarded for (i) a correct answer, (ii) an incorrect answer and (iii) no attempt, in each part of each question (or at least in each question).
- An option to display this information to the examinee next to each part of each question.
I'd like to hear from other people about their preferred MRQ frameworks.