Assessment that tells you something
Most course quizzes measure whether someone was recently exposed to a sentence. That is not the same as measuring capability.
Assessment in digital courses is usually a multiple-choice quiz at the end of a module, passed on the second attempt by nearly everyone. It generates a completion tick and almost no information.
That is a missed opportunity in two directions: the learner learns nothing about their own gaps, and the instructor learns nothing about where the course is failing.
Self-reported progress has the same weakness in research and operations; this resource explains the bias that appears when people assess their own behaviour.
Decide what the assessment is for
Three different purposes get conflated, and they call for different instruments.
- Feedback to the learner while they still have time to act on it. Frequent, low stakes, immediate.
- Evidence of capability at the end — the thing a certificate would attest to, if you issue one.
- Diagnosis of the course itself: which explanations are not working, in aggregate and anonymously.
A single end-of-module quiz serves the third purpose accidentally and the first two badly.
For broader evidence-based guidance on teaching and course design, the Carnegie Mellon Eberly Center publishes university teaching resources.
Assess the doing, not the remembering
If the outcome is a task, the assessment should be the task. Recall questions are easy to build and easy to mark, which is why they are everywhere, and they measure a proxy.
Practical assessment is more work to design and more work to evaluate. The cost is real and it is the difference between a certificate that means something and one that certifies attendance.
The useful question about any assessment is whether someone could pass it without being able to do the thing. If they could, it is measuring something else.
Self-assessment needs a rubric
At scale, instructor marking of every submission is often impossible, and self- or peer assessment becomes the practical option. It works considerably better than its reputation, but only with an explicit rubric.
'Check your work against the example' produces a vague comparison. A rubric that lists what a correct solution does, what a partially correct one misses, and the three most common errors gives a learner something they can genuinely apply to their own work.
Use the aggregate to fix the course
The most valuable output of assessment is not any individual result. It is the pattern: which question does the cohort get wrong, at which point does the failure rate jump, which exercise is abandoned most.
Those points map directly onto defects in the material. A question that eighty percent get wrong is not a hard question; it is an explanation that did not work, and it is the highest-value thing on your revision list.
Be honest about certificates
If a certificate is issued, be clear about what it attests to. 'Completed all modules' and 'demonstrated the capability in an assessed exercise' are different claims, and only the second is worth anything to anyone reading it.
Overstating this is common and is a slow-acting reputational cost, because the people who discover the gap are employers, which is precisely the audience the certificate was meant to persuade.