A standard for completion, not a recipe
Creative and technical work rarely follows one neat sequence. Learners may begin with a problem, a material, a sound, a reference or an unexpected discovery. We therefore assess the evidence that exists when a project is complete, rather than requiring one prescribed process.
For assessed work from the Maker stage onward, six evidence states must be present. Their order may vary; a facilitator may teach a workflow, but nobody is assessed on having followed one. The six are evidence functions, not submission formats — how each is shown can be adapted to the learner, the subject and the project, as long as what it evidences is there.
A cycle that opens with “identify a problem” is engineering-shaped. Presenting it as universal quietly tells every musician and artist in the room that they are working wrongly. Research on where knowledge construction actually happens in maker education locates it in the reflective work around the artefact, not in adherence to a procedure[4].
Six evidence states
The wording adapts to the subject. Made may be a robot, a digital service, a song, an artwork or a facilitated community activity. Evaluated means a technical test in one project and a critique or a rehearsal in another. The category is constant; the evidence that satisfies it is native to the subject.
Explorer work is exempt. It is unassessed by definition, run as open tinkering with learner-set goals[5]. The Definition of Done applies exactly where certification applies.
One standard, spoken in each subject’s language
Every structured programme publishes its own subject-specific reading of the standard, so the interpretation is settled in advance rather than improvised. A facilitator still explains it — in a learner’s own language, or in whatever form lands — and that is presentation, not a change to what the evidence has to show.
| Evidence state | Robotics | Music | Visual arts |
|---|---|---|---|
| Aim | What problem, and for whom | What piece, what feeling, what occasion | What idea, and what response is wanted |
| Options | Sketch two or three mechanisms | Try several melodic or rhythmic ideas | Thumbnails, references, several compositions |
| Made | Built and programmed | Composed, arranged, recorded | The work produced |
| Evaluated | Tested against spec — does it work? | Rehearsed and listened back — as intended? | The critique — does it read as intended? |
| Shared | Demonstration plus build log | Performance plus session notes | Exhibition plus process journal |
| Refined | Redesigned from test data | Arrangement revised, re-recorded | Reworked from the critique |
In arts subjects the artist’s statement is the natural Shared record. Documented portfolios are a well-precedented way of making process and learning visible across settings — and around three-quarters of surveyed makerspaces already assess in some form, so this is established practice rather than an experiment[21].
Three sources of certificate evidence
The reflection is not decoration. Experiential accounts of learning place reflection as the step that helps a learner connect a specific experience to the next situation[7] — and, just as importantly here, it is what makes that connection visible to an assessor. Without it, the evidence shows only that this particular thing was built once.
The talk-through carries a lot of weight and costs the least. It requires only a facilitator’s time, it travels across subjects better than most instruments, and it compensates for the fact that digital work is far easier to assess automatically than craft work is. It is also an efficient way to distinguish genuine understanding from a polished artefact produced with substantial help — one an equivalent accessible format can replace, where that format answers the same questions — an issue that has become more pressing, not less, as generative tools have become available.
Fair, legible assessment
- Criteria are visible before work begins, written in language learners can actually use.
- Comparable evidence is compared. A learner’s work is judged against the level rubric they enrolled at, never against the strongest person in the room.
- Documentation makes progress reviewable, so a judgement can be revisited rather than remembered.
- Assessment is separated from encouragement. Both matter; conflating them makes the first meaningless and the second suspect.
- Prior learning is recognised on evidence, not provenance — against the same prerequisites or outcomes as taught learning, wherever it was acquired. The placement check does this for entry; completion is judged against the programme’s own standard.
Attendance creates an opportunity to learn. Demonstrated, explained work shows what was learned. Only the second belongs on a certificate.
Judging quality without pretending to precision
The Definition of Done establishes that work is complete. Judging how good it is, is a separate question, and our instruments for it are coarser than a percentage mark implies.
Two published frameworks inform how the level rubrics describe increasing quality without inventing false precision. Structural accounts of learning outcomes distinguish work that handles one element, several unconnected elements, several related elements, and a generalisation beyond the task[20]. Skill-acquisition accounts describe the shift from following rules, through situational judgement, to fluent practice[18]. Revised taxonomies of educational objectives treat creating as a complex form of learning rather than a decorative one[19], which is what allows a maker programme to place it in its advanced outcomes rather than at the end of everything else.
Creativity is not one thing at one scale: published work separates it into four, from a personally meaningful insight up to field-changing contribution[23]. Almost everything a learner produces sits in the first two, and that is not a deficiency — but it means a rubric that grades against professional novelty is measuring the wrong thing.
What this does not solve
Hard results are straightforward to verify: the artefact exists, it works, it was tested. Soft results — persistence, judgement, collaboration, confidence — are genuinely hard to verify, and maker educators report this consistently.
The talk-through and reflection pair is our mitigation. It is not a proof. That flag stays on the certification system until our own delivery data says otherwise, and we would rather publish the limitation than a rubric implying a precision we do not have.
Two further items are open and named rather than glossed: the exact exit criteria at each level in each subject are still being authored alongside the curricula themselves, and whether certificates should carry external accreditation or remain internally issued has not been decided. Neither is presented here as finished.