Natural Sciences: Predict-then-verify in the lab

What you'll be able to do: Turn a practical from a procedure you followed into an experiment you ran, by committing to what will happen before it does.

The gap this closes

A student completes a practical perfectly. Correct technique, clean results, good write-up. Ask them what it demonstrated and you get the title of the practical back.

This is the procedure-model gap, and it's one of the defining failures of science education. The practical was an instruction sequence. Nothing in doing it required a model of what should happen, so no model was tested, and the results (whatever they were) confirmed nothing in particular.

Prediction closes it in about thirty seconds, because a prediction is a claim from a model. Once you've committed, the result either supports the model or it doesn't, and suddenly you're doing science rather than following a recipe.

What to predict, and when

Before you start, write down:

1. The direction. Will it go up, down, faster, slower? Which way does the equilibrium shift?
2. The rough magnitude. Double? Ten percent? An order of magnitude?
3. The shape. Linear, exponential, plateauing, sigmoid?
4. Why. One clause. This is the part that gets corrected.

Item 4 is the one that makes this a learning method rather than a guessing game, A prediction without reasoning can only be right or wrong; a prediction with reasoning can be right for the wrong reason, which is the most useful outcome available.

Across the sciences

Chemistry: rates. Predict the effect of temperature on rate, and by roughly how much. Most students predict "faster" and are surprised by how much faster, which is the collision-theory insight, arriving as a surprise rather than a statement.

Chemistry: equilibrium. Predict the direction of shift before adding anything. Le Chatelier becomes a tool rather than a phrase.

Physics: circuits. Predict the current at each point before measuring. This is where "current is used up" dies, in about ninety seconds, permanently.

Physics: mechanics. Predict the acceleration, and the direction of the friction force. Free-body errors surface as wrong predictions rather than as wrong answers three weeks later.

Biology: enzymes. Predict the shape of the rate-versus-temperature curve. Students reliably predict monotonic increase and are corrected by their own data, which is far better than being told about denaturation.

Biology: osmosis. Predict mass change in each solution, with a reason. "Water moves to where there's more solute" versus "water moves down its own potential gradient" give the same prediction here and different ones elsewhere, worth noticing.

Earth science. Predict the sediment pattern, the cooling rate, the crystal size. Grain size and cooling rate is a relationship you can predict and then see.

Astronomy. Predict where a planet will be, or what phase the moon should be in tonight given what you saw last week. Checking against the sky is free.

The reconciliation step

This is the part that makes the whole thing worth doing, and it's usually skipped because the write-up doesn't ask for it:

Here's what I predicted and why: [..]. Here's what I got: [..]. Was my reasoning wrong, or was my reasoning right and something else interfered? Those are different and I want to know which.

Three outcomes, three different lessons:

• Prediction right, reasoning right: the model works. Fine, move on.
• Prediction wrong: the model is wrong somewhere. This is the valuable case and it's the one students treat as failure.
• Prediction right, reasoning wrong: the dangerous one. You'd have scored this a success. Only stating your reasoning catches it.

And a fourth that matters in real labs: the result disagrees and the experiment was flawed. Distinguishing "my model is wrong" from "my technique was poor" is itself a scientific skill, and it's what the errors-and-improvements section of a write-up is supposed to be for.

Pitfalls

1. Predicting after seeing the first data point. That's not a prediction.
2. Predicting the outcome without the reasoning. Halves the value and hides the right-for-wrong-reason case.
3. Treating a wrong prediction as a bad practical. It's the best outcome available.
4. Blaming technique automatically. Sometimes the model is wrong. Ask which.
5. The tell: your write-up's conclusion restates the title of the practical.

Try this today

Before your next practical (or before the next worked example in a textbook) write three lines: direction, magnitude, and one clause of why.

Then reconcile afterwards. The reconciliation takes two minutes and is the only part of a practical that reliably teaches anything.