Methods · Visual, spatial & data representation

Parameter sweep

Change one variable at a time and watch the shape move, until you can predict the movement before you see it.

When to use this: While you practise
Learning impact: Strong change
Learning evidence: Credible with caveats
Where you are with the topic: Building understanding
How often to use it: Worth repeating often

When to use it

You can recite a formula but have no feel for what happens when one variable changes.
A simulation or graph tool is available and you keep watching instead of predicting.
You need intuition before you memorise another static figure.

What to ask AI

Type something like: “I'm learning [relationship]. Walk me through what happens as [parameter] increases, one step at a time. Before each step, ask me to predict the graph or outcome. Do not reveal until I answer.”

Tell it to change only one parameter at a time and wait for your prediction.

Keep the hard part on your side: you predict before you see the next frame.

How it helps

Static graphs hide motion. A parameter sweep makes the relationship move, so you feel covariation instead of only naming it.

The learning happens in the gap between your prediction and the reveal, not in watching a smooth animation end to end.

Go to the extremes. That is usually where the model's structure shows, and where a wrong intuition fails first.

How to run it

Name the relationship and the single parameter you will vary.
State your starting intuition in one sentence before any sweep.
Predict the next step before each change: what moves, which way, roughly how much.
Only then look at the result and update your rule.
After several steps, predict a case you have not seen yet.
Close by stating the rule in your own words without looking at the graph.

Example exchange

You

I'm trying to get intuition for demand curves. If income rises for a normal good, what actually moves?

Start at today's equilibrium price. Income rises a little. Does quantity demanded at that price rise, fall, or stay put?

You

It should rise. People can buy more at the same price, so the demand curve shifts right.

Rightward shift. Now predict the new equilibrium price and quantity before I describe them.

Copyable prompt

I'm learning [relationship/model]. Walk me through what happens as [parameter]
increases, one step at a time. Before each step, ask me to predict. Do not reveal
until I answer. End by asking me to predict a case I haven't seen.

The Tell

Here is how you know this method has flipped: you can describe what happened in the chat and cannot predict a case you have not seen.

Watching the sweep without predicting is entertainment dressed as practice.

Force a prediction before every reveal, especially at the extremes.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Varying one parameter with prediction builds covariation and functional reasoning, intuition before formal statement, a familiar practice in maths and science teaching with dynamic graphs.

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

AI can narrate a sweep well or invent impossible curves. Treat delivery as speculative: your predictions and a trusted graph tool matter more than a fluent story about what “would” happen.