Science Education Systems · Article 43. Maya, Jia Jun, Hana and Ethan remain fictional Punggol learners. This article follows the simulation layer: how Science builds controlled artificial worlds to explore what may happen when real experiments are difficult, slow, expensive or impossible.
The 50-second parent route
A simulation is a model that runs.
It takes assumptions, rules, inputs and initial conditions, then generates outcomes.
The route is:
question → model → assumptions → parameters → initial conditions → simulation → output → comparison with reality → validation → sensitivity test → revision
The key question is:
What part of reality is the simulation representing, and what did we leave out?
A simulation can produce valuable scientific insight.
It does not become empirical evidence merely because the graphics look realistic.
This article extends How Scientific Models Grow With the Learner, How Scientific Probability Works and How Scientific Validation Works.
1. Simulation is useful when direct experimentation is hard
Planetary motion.
weather.
epidemics.
ecosystems.
aircraft flow.
molecular interactions.
Some systems are too large, too small, too slow, too dangerous or too expensive to manipulate freely.
2. A simulation is not reality
It is a rule-governed representation.
The output is only as meaningful as the model, assumptions and inputs that generate it.
3. Simulations make assumptions executable
A written model says how variables relate.
A simulation lets those relationships evolve across time or repeated scenarios.
This makes hidden consequences easier to inspect.
4. Initial conditions matter
Where does the system begin?
Temperature.
population.
velocity.
concentration.
location.
Different starting states can produce different outcomes under the same rules.
5. Parameters define the model’s behaviour
Growth rate.
reaction rate.
friction coefficient.
transmission probability.
Parameters compress scientific relationships into values the model can use.
6. Maya’s simulation error is animation realism
The simulation looks like a real laboratory.
She assumes the output must therefore be realistic.
Her repair:
inspect the equations and assumptions behind the picture.
7. Jia Jun’s simulation error is parameter blindness
He changes an input and sees the output move.
He forgets that the chosen parameter values may be uncertain.
His repair:
test a plausible range, not one magic number.
8. Hana’s simulation error is rejecting simulated evidence completely
She says:
“It is not real, so it is useless.”
Her repair:
simulations can test implications of models, compare scenarios and guide experiments when interpreted correctly.
9. Ethan’s simulation error is unlimited scenario generation
He runs thousands of scenarios without a scientific question.
His repair:
define what uncertainty or mechanism each run is testing.
10. Primary Science can use simulations as controlled representations
A virtual circuit.
a life-cycle animation.
a shadow model.
a water-cycle simulation.
Students can change one condition and observe the model’s response.
11. Simulation should not replace concrete experience automatically
A child who has never built a real circuit may treat perfect virtual wires as normal.
Real components introduce loose connections, material limits and measurement noise.
Physical experience reveals friction the simulation may omit.
12. The strongest learning sequence often alternates
real observation → model → simulation → prediction → real test → model revision.
This keeps the simulated world connected to evidence.
13. Simulations support counterfactuals
What if rainfall decreases?
what if resistance doubles?
what if population growth slows?
what if one component fails?
We can explore alternative worlds without physically creating all of them.
14. Simulations support sensitivity analysis
Change one parameter slightly.
Does the output change slightly or dramatically?
This reveals which assumptions control the result most strongly.
15. Simulations support stress testing
Extreme heat.
high load.
rare failure.
unusual traffic.
Scientists and engineers can test difficult conditions more safely in models before real deployment.
16. Simulation and systems thinking are closely linked
Inputs.
stocks.
flows.
feedback.
delays.
Simulations let these interactions unfold dynamically.
17. Feedback can create surprising simulation outcomes
A small change amplifies.
another stabilises.
a delayed response overshoots.
Simulation helps learners see why intuition can fail in complex systems.
18. Simulation and probability are closely linked
Probabilistic models can generate many possible futures.
Repeated runs reveal distributions rather than one deterministic path.
19. Monte Carlo simulation samples uncertainty
Choose plausible parameter values according to a probability model.
Run many scenarios.
Observe the output distribution.
This estimates risk when direct calculation is difficult.
20. One simulation run proves very little in a probabilistic model
A rare outcome can occur in one run.
The important question is how frequently outcomes appear across many runs.
21. Simulation output should carry units and meaning
Numbers without interpretation are not scientific understanding.
What quantity does each output represent?
What unit?
What time step?
What spatial resolution?
22. Resolution creates computational trade-offs
Finer resolution can capture more detail.
It also increases computational cost.
Scientists choose a level that preserves the important dynamics for the question.
23. Time step matters
If the simulation updates too coarsely, rapid changes may be missed.
If it updates extremely finely, computation may become expensive.
Numerical choices can affect results.
24. Numerical error is a simulation uncertainty
Computers approximate some equations step by step.
The approximation method itself can introduce error.
Verification checks whether the code solves the intended mathematical model correctly.
25. Validation is different
Verification asks:
Did we implement the model correctly?
Validation asks:
Does the model represent reality well enough for the intended use?
See How Scientific Validation Works.
26. A perfect simulation of the wrong model is still wrong for the job
Code can run flawlessly while the assumptions omit a critical mechanism.
Computational precision does not rescue conceptual error.
27. Simulations need empirical anchors
Parameters can come from experiments.
outputs can be compared with observations.
predictions can be tested later.
Without empirical contact, the simulation risks becoming internally consistent storytelling.
28. Calibration can fit a simulation to known data
Adjust parameter values so the model reproduces observed behaviour.
But fitting alone does not prove generalisation.
New data should test the model independently.
29. Overfitting can occur in simulations too
A complex model can be tuned to reproduce historical data beautifully and still predict new cases poorly.
Validation on unseen evidence remains essential.
30. Multiple models can fit the same data
If different mechanisms generate similar observed patterns, the data may not discriminate among them.
Simulation should reveal this uncertainty rather than hide it.
31. Ensemble modelling compares multiple simulations
Different models.
different parameter choices.
different initial conditions.
If conclusions converge, confidence can increase.
If they diverge, the disagreement maps uncertainty.
32. Weather forecasting uses ensembles for this reason
Small uncertainties in initial atmospheric conditions can produce different future paths.
Multiple runs help estimate the range of plausible outcomes.
33. Climate models operate at a different prediction scale
They are used to understand long-term statistical patterns and system responses, not to predict the exact weather on one specific distant afternoon.
Scale defines the validation target.
34. Molecular simulations explore invisible scales
They can model interactions that are difficult to observe directly.
But the force models and approximations still require validation.
35. Engineering simulations can reduce development risk
Airflow.
structural load.
heat.
fluid movement.
Simulations help identify likely failures before expensive prototypes are built.
36. Simulations do not eliminate real testing in safety-critical systems
Physical tests remain important because reality contains manufacturing variation, unforeseen interactions and model limitations.
Simulation and experiment complement one another.
37. Simulation supports ethical Science
Some dangerous, invasive or harmful experiments can be reduced or avoided when validated simulations provide sufficient information.
Simulation can be a responsible alternative when used within its limits.
38. Simulation can also create ethical risks
A model may encode biased data.
policy decisions may rely too heavily on uncertain outputs.
users may mistake simulated scenarios for observed facts.
Transparent assumptions matter.
39. AI systems are themselves models that can simulate language and behaviour
They generate plausible continuations rather than directly measuring the world.
Fluency should not be mistaken for empirical observation.
40. AI can generate synthetic data
Synthetic data can be useful for training, privacy protection or scenario testing.
But it should be labelled clearly.
Generated data is not a replacement for real evidence when the scientific question requires empirical observation.
41. AI can help build simple educational simulations
Useful prompts:
“Create a simple population model with adjustable growth rate.”
“Let me vary one parameter and predict before showing the outcome.”
“Explain which assumptions make the simulation unrealistic.”
“Give me one result that would fail validation against real data.”
42. Parents can use simulations as question generators
After a child uses a virtual experiment, ask:
“What would be messier in real life?”
“Which variables did the simulation keep perfect?”
“How could we test one part physically?”
43. Small-group tuition can compare model and reality
Run a virtual circuit.
Predict the ideal result.
Build the physical circuit.
Compare.
Differences reveal resistance, connection quality and measurement limitations.
44. A compact simulation checklist
- What scientific question is the simulation answering?
- What real system does it represent?
- What assumptions are built in?
- What parameters control behaviour?
- What initial conditions are used?
- What scale and resolution are represented?
- What important processes are omitted?
- Has the implementation been verified?
- Has the model been validated against independent evidence?
- How sensitive are outputs to assumptions?
- What uncertainty should accompany the result?
- What real observation could test the simulation?
45. Frequently asked questions
What is a scientific simulation?
A scientific simulation is an executable model that generates system behaviour from stated rules, parameters and initial conditions.
Is simulation data the same as experimental data?
No. Simulation output is generated by a model; experimental data is produced by observing or measuring the real world.
Why are simulations useful?
They allow scientists to explore complex, large, small, slow, expensive or dangerous systems and compare many scenarios efficiently.
What is validation?
Validation checks whether the simulation represents the real system well enough for its intended purpose.
How do simulations help PSLE Science?
They help learners visualise systems, vary one condition, make predictions and compare model outcomes with real observations.
How do simulations change in Secondary Science?
They become more quantitative through parameters, numerical models, probability, systems dynamics and comparison with measured data.
46. Continue the Science Education Systems series
Conclusion: A simulation is a question asked inside a model
Maya sees the animation.
Jia Jun sees the parameters.
Hana sees the assumptions.
Ethan sees the possible worlds.
Science needs all four.
Build the model.
run it.
stress it.
compare it with reality.
keep what survives.
revise what fails.
The simulation becomes scientifically valuable when it remains accountable to the world outside the screen.
