measuring learning

In university, I wrote an honours thesis on locating earthquakes.  To solve the problem of locating earthquakes we must use indirect observations – our measurements of seismic waves – to learn about where an earthquake occurred and the size of the earthquake. 

This is very like measuring learning.  It's not possible to directly measure whether someone has learnt something or not.  At best, we try to track imperfect measures like skill development, recall of facts or a behaviour change to show learning progress.  These measures give you little useful information without knowledge of the learner.  Take a spelling test, this gives us next to no information about learning unless we know whether the learner could spell the word prior to the test, even then, we don't know that the learner will be able to spell the word next week or use the work correctly in writing.

When you use indirect observations to find the location of an earthquake you must be aware of the pieces of information you don't know.  For example, you don't know exactly which types of rock the seismic waves passed through.  You predict this based on a model.

Measuring learning is the same.  It is based on models and assumptions.  It is wise to remember what those models and assumptions are when you attempt to measure learning.