I’ve been dabbling in the automatic marking of tests and quizzes for several years now. By this, I mean a web-based set of questions that a student completes on a specific topic, that automatically grades the answers as correct/incorrect (and sometimes gives partial credit) and returns the mark (sometimes with feedback) to the student at a specific time. Before you think – oh this is good, no marking, I’d warn you about setup burden.
What kinds of assignments make really good auto-tests? I have used them for the following:
- pre-laboratory exercises that give practice at calculations, some safety aspects, and identifying products/balancing equations (Blackboard Test/Pool)
online safety quiz (Blackboard Test/Pool)
assessed tutorial with a very tight deadline before the exam (Blackboard Test/Pool)
referencing and academic conduct test (Blackboard Test/Pool)
diagnostic test (new for 2017! Google Form Test)
The technology has limitations, particularly related to the type of questions you can ask. I find the following types useful:
- multiple choice questions
calculated numeric [with the caveat that Blackboard can’t deal with 5.5 and 5,5, units, or number of decimal places]
fill in the blank or short answer [with the caveat that students often can’t spell (even when given a list of words to select their answer from), and sometimes deducing the syntax of the required answer is tricky]
matching pairs [really good for reactant/product equation matching in transition metal redox chemistry]
I also like the ability to write a pool of questions and a system that allows each student to be asked a number of questions from the pool. If every question is from a different pool, this reduces the scope for collusion. An example of a good pool question stem for a calculated numeric question:
Calculate the mass of copper sulfate required to prepare a [Y] molar aqueous solution in a [X] mL volumetric flask.
You can see how simple it is to vary X and Y within the realms of possibility, generate all the correct answers in excel and make a pool of questions.
The setup burden is how long it takes to create the initial test. As a rough guide, I’d estimate it to be at least twice as long as it would take to manually mark! So for a pre-lab done by 50 students, taking me 10 hours to mark, I’d expect to spend about 10 hours developing the online version. I do not recommend doing the online test thing unless you know you can use it for at least 2 years – one reason for doing it is to reduce the marking load and you don’t really start to make gains until the 3rd year of running. On the other hand, it’s a convenient way to ship time from semester (marking time) into quieter times of the year (prep time). I estimate that each test requires 1 – 2 hours of tweaking and set-up each year, usually after reviewing the analytics from Blackboard, weeding out poorer questions, adding a couple of new ones…that sort of thing.
Why do I do this? Well each of the assignments I’ve outlined are reason enough in themselves, but some have transitioned from paper-based to online (pre-labs, diagnostic test) and some would not exist if they could not be online (safety, referencing, academic conduct, assessed tutorial). So sometimes there is no reduction in marking time for me because I wouldn’t offer the assignment in an alternative manner. Technology facilitates the use of formative tests to aid learning, so I use it.
This year I’m expanding my range of formative tests to transfer my 1st year spectroscopy ‘drill’ questions into an online format. When teaching things like the equations of light, basic NMR, IR etc, I recognize the value in doing lots of examples. I also recognize the value in those examples stepping up in difficulty every so often, I’ve been calling them levels.
For example, using the equation E = hν
Level 1 – calculation of E in J, with ν in Hz
Level 2 – calculation of E in kJ/mol
Level 3 – calculation of E in kJ/mol with ν in ‘insert standard prefix here’ Hz
Level 4 – calculation of ν with energy in J
Level 5 – …
You get the idea anyway. I read a paper on this a few years back, about stepping up calculations in small steps. So I’m making question pools for each level, bundling a few levels together into a quiz then setting a minimum score requirement to gain access to the next levels. Students will do quiz 1 and if their mark is high enough (80%+) they get access to quiz 2. If it isn’t, they’ll get access to a couple of worked examples and the chance to re-do quiz 1 to get the mark.
I’m aware that this type of drill enforces a purely algorithmic approach, but if my students can’t do these bits, they are going to run into a whole lot of problems at higher levels. When setting exam questions, I balance the question between the algorithmic problem solving stuff like performing a calculation, and the ‘explain your answer’ part where they need to demonstrate a degree of understanding. We can argue over the appropriate balance between those sections but I think the algorithmic stuff should be 40 – 60% of the marks available (depending on the level of the paper) and the balance should be the explanation stuff, or higher level problem solving such as unseen, unfamiliar problem types. With this balance I’m saying ‘well you can probably pass the exam if you can do the basics but you need to show more to get a great mark’. I also assume that intended learning outcomes define a passing mark (40%) or a low 2:2 mark (50%), rather than a 100% mark.
The experience of setting up and running diagnostic test through Google Form Tests requires a post on it’s own so I’ll come back to that.