Three broad elements on which every assessment should rest are described by
the Assessment Triangle from the National Research Council’s report Knowing
What Students Know [NRC 2001] shown in Figure 1.
According to Knowing What Students Know, an effective assessment design
requires:
a model of student cognition and learning in the field of study;
well-designed and tested assessment questions and tasks, often called items;
and ways to make inferences about student competence for the particular
context of use. (p. 296)
These elements are of course inextricably linked, and reflect concerns similar to
those addressed in the conception of constructive alignment [Biggs 1999], regarding
the desirability of achieving goodness-of-fit among learning outcomes,
instructional approach, and assessment.
Models of student learning should specify the most important aspects of student
achievement to assess, and they provide clues about the types of tasks that
will elicit evidence and the types of inferences that can connect observations
to learning models and ideas about cognition. To collect responses that serve
as high-quality evidence, items themselves need to be systematically developed
with both the learning model and the character of subsequent inferences in mind,
and they need to be trialed, and the results of the trials systematically examined.
Finally, the nature of inferences desired provides the “why” of it all — if we
don’t know what we want to do with the assessment information, then we can’t
figure out what the student model or the items should be. Of course, context
determines many specifics of the assessment.