Formative assessment of learning and experience
You find out what went wrong in January, from the course evaluations
Half the room never understood the thing week five depended on, and the exam is where you find out.
What it is
Two different things the fellowship treats together, because both are about knowing where students are before it is too late to act.
Assessment of learning is a short ungraded check on whether a session landed. Assessment of experience is a short confidential survey on how the course feels to be in. They use different instruments and answer different questions, and the fellowship’s own protocol keeps them separate.
Techniques
Exit tickets
Pick one, run it at the end of a session, read the answers the same night, open the next session with what you found.
- Muddiest point. What was the muddiest point in today’s session?
- Minute paper. What was the most important thing you learned, and what important question remains unanswered?
- Confidence rating. One to five, how confident are you that you could explain today’s core concept to a first-year student? This one reaches the quiet students, because it asks for a number rather than an admission.
Keep it ungraded and say so. The moment a quick check is graded it becomes a performance and stops being diagnostic.
Opening the next session
“Eleven of you named the same muddiest point last week, so we are starting there rather than where the schedule says.”
The collecting is not the practice. Students work out within two weeks whether the answers go anywhere, and if they do not, response quality collapses.
The student experience survey
- Find an outside collector first. A colleague in another department owns the link and returns you an aggregate, reciprocated. It is the load-bearing step, because you asking directly is a different instrument.
- Pick four or five of the seven learning conditions, not all seven. The protocol says this itself, and it is the first defence against survey fatigue.
- Tell students four things. You will use it to improve their environment, you will ask again with the same questions, responses are confidential and collected outside the department, and you will debrief the results.
- Give the last five minutes of a session and leave it open two days.
- Debrief within a week. One stop, one start, one continue, and one thing you are not changing.
- Ask the same items next time. Changing the wording forfeits the only comparison you have.
The debrief
“Here is what you told me, including the uncomfortable parts. Here is what I am starting because you said it mattered. Here is what I am stopping. Here is what I am continuing, because most of you said it was working. And here is one thing I am not changing, and why.”
In my courses
Run in EPI 501 and EPI 553. Planned for EPI 601, even weeks, with surveys at weeks 4 and 10.
I will administer the E3 student experience survey at two points (weeks 4 and 10). I will use informal formative assessments weekly: exit tickets after lectures on complex topics (e.g., ‘What is still unclear about interaction on the additive vs. multiplicative scale?’), self-rating checks, and quick writes.
The exit tickets are targeted at the three concept-heavy sessions: confounding, misclassification, and interaction. The benchmark comes from EPI 553, where 88 percent of students agreed they received clear academic feedback; EPI 501 surveys had shown students wanting more. Used before the fellowship.
Evidence
The two halves have very different support, and running them together does not average out.
The muddiest point has never been tested against a control anywhere. The minute paper has two quasi-experiments from 1998. The fellowship named these and supplied no prompt wording, timing, or collection method.
The survey instrument is worth flagging: no peer-reviewed validation, no reported alphas, and no factor structure could be located for the seven learning conditions (PERTS / Student Experience Project, 2022). Treat any description of it as psychometrically validated as unverified.
One general caveat. A meta-analysis of feedback interventions found a substantial minority actually reduced performance (Kluger & DeNisi, 1996). The corpus is industrial rather than classroom, so treat it qualitatively, but it corrects the assumption that more feedback is always better.
With twelve students, report the numerator over the denominator and nothing else. Seven of twelve is a fact. Fifty-eight percent is a claim the sample cannot support.