Tuesday, July 24, 2007

Idea of the Day: Multiple Choice Tests Without System Gaming

One of the oddities of multiple choice tests is that there is typically one and only one correct answer. So if you can determine that the other answers are wrong, or that some answer is wrong, you will do better on the test. However, this does not accurately measure a person's knowledge. One could in theory pass a test without knowing any of the correct answers, but a sufficiently large number of wrong answers. Secondly one can easily fail almost completely with still having some knowledge on the subject. I propose that tests should be designed without this requirement of one and only one correct answer. Here's an example:

1) 4 * 5 =
a) 12
b) 15
c) 30
d) 9

2) A banana is,
a) a fruit
b) a tree
c) a card game
d) a type of seat

3) The war of 1812 took place in,
a) 1811
b) 1812
c) 1814
d) 1816

This results in each question either bubbled or not bubbled being equal to a true or false test. So each question, rather than finding the best answer you answer four true or false questions about the given situation. Then it's a simply process of giving -.5 for each wrong answer and .5 for each correct answer.


Warren said...

Uh ... you forgot "none of the above" for #1.


Tatarize said...

Oddly, no I didn't. Bubble nothing. That is the correct answer. Since none of them apply.

Kristina said...

Soooo...there is no correct answer for number one; both a) and d) are correct for number 2; and b) and c) are correct for number 3. I have to agree that this method would more accurately measure a person's knowledge, but I object on the basis that I am really good at multiple choice tests.