A lab day exposes the gap between your fastest and slowest groups faster than almost any other activity, because the procedure itself is a bottleneck. A group that's shaky on the steps burns the whole period just getting the setup right and never reaches the actual analysis question. A group that's confident is done with time to spare and starts drifting. One handout for the whole room can't serve both.
The fix isn't four different labs — it's one lab with a menu of scaffolds around the same procedure, so every group reaches the same core skill by a different amount of support getting there.
Scaffold the procedure, not the science
The actual skill you're assessing — reading a meniscus, identifying an independent variable, calculating percent error — should be identical across every version. What changes is how much of the setup and sequencing is already done for them.
- Full scaffold. Numbered steps, a partially filled-in data table, and a worked example of the first data point already calculated.
- Light scaffold. A numbered procedure but a blank data table, and the calculation method described but not modeled.
- Open. A goal and available materials, with the procedure sequence left for the group to plan themselves before you check it off.
All three groups collect and analyze the same kind of data and answer the same core analysis question at the end — the ramp is in how much of the "how do I even start" work is done for them going in.
Getting the three versions written fast
Writing a fully scaffolded version, a light version, and an open version of the same lab by hand is a lot of duplicate effort for one afternoon. This is where AI earns its keep, since the underlying procedure doesn't change — only the amount shown.
Try this:
- "Here's my lab procedure: [paste it]. Write three versions: a fully scaffolded version with numbered steps and a partially completed data table, a light version with numbered steps but a blank data table, and an open version that gives the goal and materials but leaves the procedure sequence for students to plan. All three should lead to the same data and the same final analysis question."
- "Write one worked example row for the data table, showing the full calculation for the first data point, for the fully scaffolded version only."
Check that the analysis question stays identical across all three drafts — AI sometimes simplifies the open version's final question along with the procedure, which defeats the point of keeping the actual skill constant.
Assigning without making it obvious
Hand out versions by table number or a quick pre-lab check-in rather than announcing which version is "the hard one." A fast informal check the day before — can this group state the procedure back to you in order — tells you who needs the full scaffold without a test.
Our [differentiation plan generator](/differentiation-plan) can build this three-tier scaffold structure for a specific lab from the standard down, if you want the full ramp planned out rather than assembled from a single procedure. And for the lab itself, our [STEM activity generator](/stem-activity) is the faster starting point when you're building the procedure from scratch rather than adapting one you already run every year.
What still needs your judgment
- Which group gets which version. Base it on where each group actually is with the specific skill this lab needs, not a fixed tier that follows students all semester.
- When the open version is a bad fit. A first-time procedure with real safety steps isn't the moment for an unscaffolded version — save the open tier for procedures students have some baseline comfort with.
- Whether the analysis question stayed fair. Read all three versions' final questions side by side before lab day — if the open version's question is quietly easier, the tiering isn't doing what you think it's doing.
