In our Taxonomy of Mark Schemes we noted that exam questions usually
assess either the quality of students'
'work or how correct their answers are
to questions of varying difficulty.
We have developed a third approach, in which students are judged in
terms of how much help they need in
order to produce an acceptable
response to a complex question. With
this model everyone succeeds, even if
they need a lot of support. In our
trials it has proved very popular with
students. Through a computer
interface, a student's early attempts
at an answer are evaluated and prompts
are given to improve the answer until
it is considered good enough: the
amount of help given is a measure of
the student's understanding.
The model is not complete yet; improved programming is needed to automate the prompting process and further research is needed to optimise the scoring system.