Locks and checks help students learn. They are not integrity controls, and this page says plainly why, so you can plan assessment around it. Everything here is true in every edition.
What they do well
- Keep a class on the same task. Locked inputs and outputs mean every hand-in reads the same data and gives the same kind of answer, so the grader can compare them.
- Give feedback at once. Check my work answers in a moment, with the lab’s own words, so a student fixes a mistake while it is fresh.
- Mark the same way twice. The checks are rules, not judgments: Check my work and
spl gradegive the same marks, on any computer. - Notice a changed lock. The grader compares each hand-in’s locked blocks with the starter’s fingerprint, so a lock changed outside the app loses its points.
What they can’t do
- Students control their own files. A project file is a zip of plain files. A student who wants to can change anything in it, outside the app.
- A valid model can come from elsewhere. The file format is published, so a classmate, a website or an AI assistant can write a model that passes every check.
- Checks don’t see understanding. A model that passes is a model that meets the rules, not proof that the student knows why.
- The session summary is evidence, not proof. Its times and counts come from the student’s own computer, so read them as a picture of how the work went, never as marks.
Assessment that holds
- Grade reasoning in the room. Ask students to explain an error code, predict a shape, or say why a change helped, during the lab or in a short checkpoint.
- Ask questions the model can’t answer. The worksheets’ questions (“why does the training accuracy reach 100% while the test accuracy doesn’t?”) are where understanding shows.
- Use open-ended work. A project on data the student chose is harder to copy, and its choices are worth talking through.
- Turn the console off for an assessment, with the lab’s policy, if you want the blocks alone.
- Talk with the student when a hand-in surprises you. Open their file in the app: its model, its settings and, when they kept them, its saved versions show how the work grew.
Hidden checks
Hidden checks are an Education feature, coming in a later release with the lab editor: a check kept out of the student’s file and marked from your grading pack. They make a lab harder to game, since a student can’t read the check, but the same limits hold: a model written elsewhere can still pass them.