This is specifically from the perspective of math education, but may apply to other areas just as well.

Why do we give assessments? Surely the answer is to find out what students know and what students can do. Not in general; we want to know specifically whether students know certain things. We have learning targets and standards. However, a common problem I see with assessments is that they don’t provide this information. Looking at it in terms of data, we want to isolate the variables we’re assessing as best we can and eliminate confounding variables.

How do assessments fail in this regard? First, there is question design. If a question on a geometry test is about scale factors, it should not require the student know how to convert from feet to miles unless we are specifically intending to assess this knowledge. Even then, a student getting the wrong answer does not tell you what aspect of the question the student does or doesn’t understand. One solution to this is to make different parts of the question count for different things, possibly even distributing the question’s points over multiple standards or learning targets. For example, the question just described could have 1 point under the “measurement and data” standard and 3 points under the “geometric similarity” standard. This has the drawback of making grading a lot more painstaking, and can be impractical to implement.

Second, there is the way in which assessments are given. A test should not be timed, for example, unless speed is specifically being assessed. Even the ability to take a written test in a room full of other students varies, and again, unless this ability is specifically and intentionally being assessed, then its factoring into students’ scores is confounding and gives inaccurate results. For students with IEPs, 504 plans, etc., there is often an option to take a test in a separate room. However, according to the principles of Universal Design this should not present an issue for any students, plan or no. Here again there are practical considerations, namely it is impractical or even physically impossible for a large number of students to be taking tests in different rooms. One alternative is to assess students in a different way, possibly incorporating student choice. Traditional written assessments are not the be-all end-all.

Part of addressing these issues of course is not only assessment design, but also addressing wider problems within education. Effective assessment requires highly qualified teachers, reasonable class sizes, availability of material resources, and so on.