PSLE Science Reality Lab Vol No.016 | “The Simulation Shows It” — Is a Model Prediction the Same as an Observation?

Series ID: PSLE-SCI-REALITY-0016

Wait, What? A Beautiful Animation Can Show Something That Has Never Happened

A screen shows water flowing through a city, clouds moving across a planet, heat spreading through a building or a spacecraft landing safely. The movement looks precise. The colours look scientific. Numbers update every second.

But what are you actually looking at?

Sometimes you are looking at observations. Sometimes you are looking at a reconstruction built from observations. Sometimes you are looking at a computer simulation: a model running rules and calculations to produce an output. A simulation can be extremely useful without being a direct observation of the real world.

The scientific skill is not to dismiss simulations. It is to know what kind of evidence they are.

Quick Answer

When someone says “the simulation shows…”, ask:

What real system is being represented?
Which inputs came from observations or measurements?
Which rules or relationships were built into the model?
What assumptions simplify the real world?
Which output was calculated rather than directly measured?
Has the model been checked against observations it did not simply copy?
Does it work only within a particular range or condition?
What would count as evidence that the model needs revision?

The quiet rule is: model output is evidence about what the model predicts; observations are evidence about what the real system did. Strong science connects the two carefully.

The Owned Learner Job

This Reality Lab owns one transfer job: how to read a scientific claim presented through a computer simulation without treating the simulated output as though it were automatically a direct observation.

It does not replace existing eduKateSengkang guides on scientific models, observations, predictions or evidence. It applies those owners to a real communication object that students increasingly meet in videos, news graphics, apps and interactive websites.

Reality Lab Case: The Virtual Pond

Imagine a fictional computer model of a pond. The model starts with water temperature, light level and the number of a certain water plant. It then calculates how the plant population might change over several weeks.

The screen shows the plant population rising quickly when the light level is increased. A student says, “The computer observed that more light causes this plant population to increase.”

The wording is wrong in an important way. The computer did not observe the pond. It calculated an outcome from the model’s inputs and rules.

A stronger statement is: “Under the rules and assumptions of this model, increasing the light input produced a higher simulated plant population.”

Now the statement tells us where the result came from.

Four Layers: Observation → Input → Model → Output

A useful learner map is:

Observation: something measured or recorded from the real system.
Input: information supplied to the model.
Model: the rules, relationships and simplifying assumptions used to calculate what happens next.
Output: the result produced by the model.

These layers can interact. A model may use observations as inputs. Scientists may compare outputs with later observations. They may change the model when it repeatedly fails. But keeping the layers separate prevents one of the most common reasoning errors: treating a calculated picture as though a camera recorded it directly.

Models Are Useful Because Reality Is Difficult

Some systems are too large, too slow, too dangerous, too expensive or simply impossible to manipulate directly. We do not have a spare Earth for climate experiments. We cannot repeatedly crash full-size spacecraft just to explore every design choice. We cannot wait centuries whenever we want to study a long-term process.

NASA explains climate models as laboratories in computers: scientists can change a factor in the model and study how the represented system responds. That makes simulation powerful. It does not remove the need for observations. It makes the relationship between model and observation more important.

A Model Contains Decisions

Even when a model is based on good science, someone has decided:

which parts of the system to include;
which parts to leave out;
which relationships matter;
how finely to divide space or time;
which measured values to use as inputs;
how uncertain inputs will be handled;
where the model should stop being trusted.

Those decisions do not make a model unscientific. They make its assumptions part of the evidence boundary.

Worked Case 2: A Flood Animation

A fictional animation shows water spreading across a neighbourhood after heavy rain. The caption says, “This is what will happen during a major storm.”

Before accepting the sentence literally, ask what the simulation used:

How much rain?
Over what time?
What drainage capacity?
What ground height data?
Were blocked drains included?
Was water able to flow through buildings or walls unrealistically?
Was the model checked against previous real floods?

The animation can still be valuable. The better wording might be: “This simulation shows one modelled flood outcome under the stated rainfall and drainage assumptions.”

That statement respects both the usefulness and the limits of the model.

Prediction Is a Test Opportunity

A good model often earns trust by making predictions that survive comparison with later or separate observations. That does not make the model permanently correct. It shows that the model has worked under the conditions tested so far.

Suppose a model predicts the temperature at ten locations. Later measurements match closely at eight locations but differ strongly at two. Scientists do not need to choose between “the model is perfect” and “the model is useless”. They can ask why the mismatch occurred.

Were the measurements reliable?
Was an important local factor missing?
Was the model’s spatial resolution too coarse?
Was one input wrong?
Does the mismatch reveal a process that needs better representation?

This is model-building as a scientific enterprise: prediction, comparison, revision and return.

Do Not Confuse Realistic Appearance With Scientific Accuracy

A simulation can look extraordinarily realistic because of smooth animation, detailed textures and carefully chosen colours. None of those visual features prove that the underlying scientific model is accurate.

The reverse is also true. A simple graph or crude diagram can come from a well-tested model. Scientific quality is not measured by how cinematic the output looks.

What Would Strengthen a Simulation-Based Claim?

The model’s purpose is stated clearly.
Important inputs and assumptions are disclosed.
The output quantity is defined.
The model is checked against independent observations where possible.
Predictions are tested rather than only fitted to known examples.
The model’s valid range and limits are stated.
Alternative models or parameter choices are considered where relevant.
Mismatch with observations is reported rather than hidden.

What Would Weaken the Reader’s Confidence?

The simulation is presented as if it were a direct recording.
Inputs or assumptions are not described.
The model is judged only by how realistic the animation looks.
The same observations used to build the model are the only evidence offered to “prove” it.
A prediction far outside the tested range is presented with no uncertainty.
Disagreement with real observations is ignored.

Counterexample: Sometimes the Simulation Is the Right Tool

A learner might overcorrect and say, “If it is only a simulation, it is not evidence.” That is too simple.

Simulation results can be powerful scientific evidence about the consequences of a model. They can reveal patterns, test mechanisms, compare scenarios and generate predictions. Scientists may combine simulations with measurements to study systems that cannot be fully observed or manipulated.

The key is to name the evidence correctly. A simulation result tells us what follows from the model under its inputs and assumptions. Confidence about reality grows when that model is grounded in scientific mechanisms and survives comparison with observations.

PSLE-Style Transfer Case

A computer model predicts how the temperature of a closed box changes when the thickness of insulation is increased. The model predicts that thicker insulation produces a slower temperature change. A student writes, “The experiment proves that thicker insulation always slows temperature change.”

Earliest weak link: no real experiment has been described. The evidence given is a model prediction.

Improved statement: “In the model, increasing insulation thickness produced a slower simulated temperature change. Measurements from a real set-up would be needed to test how well that prediction matches the actual system.”

The learner has not rejected the model. The learner has classified the evidence correctly.

Delayed Independent Return

Next time you watch a scientific animation, pause and write four words:

observed?
input?
calculated?
validated?

If you cannot tell which is which, the communication may still be useful, but you do not yet know what kind of scientific claim it supports.

Useful eduKateSengkang Routes

Parent and Tutor Teaching Guide

Use weather maps, simple online animations or an invented simulation screenshot. Ask the learner to label each visible element as observed data, model input, calculated output or unknown. Do not begin by telling the learner whether the model is “good”. First make the evidence layers visible.

Then ask the most powerful follow-up question: “What real observation could check this prediction?” That moves the learner from passive viewing to scientific testing.

Finally, give a case in which the model and observation disagree. A mature learner should not automatically throw away either one. The learner should consider measurement error, missing variables, incorrect assumptions, scale and limits. That is a much stronger habit than treating computer output as authority.

Authoritative Sources

The Quiet Rule to Keep

A simulation can help us see what a model implies. Reality gets the final vote.

Small Group Tutorials