Final Assessment

AI for Science Teachers — Final Assessment

Thirty-five questions across the orientation and eight modules — including scenario, error-identification, graph-interpretation, safety, privacy and prompt-evaluation items. Pass mark 75%. Attempt again after focused revision if needed.

1. Generative AI produces fluent text by predicting likely words. What does this mean for a science teacher?

It can be confidently wrong; verify every output.It is always scientifically accurate.It understands science like an expert.It never needs review.

2. Which may never be entered into a public AI tool?

Identifiable student information.A clearly labelled synthetic dataset.A public curriculum topic.A fictional example scenario.

3. How must a synthetic dataset be presented?

Clearly labelled as synthetic, never as measured data.As real measured data to look convincing.With no label at all.As peer-reviewed evidence.

4. What is AI's appropriate role in scientific inquiry?

A brainstorming and feedback tool, not a source of truth.The final authority on scientific facts.A replacement for observation and evidence.A grader of student experiments.

5. Which is a testable investigable question?

'How does salt amount affect ice-melting time?''Is science interesting?''What is the best experiment ever?''Should we like chemistry?'

6. What is the difference between a hypothesis and a prediction?

A hypothesis is a testable explanation; a prediction is the expected outcome.They are identical.A prediction needs no evidence.A hypothesis is always correct.

7. When is a virtual laboratory a better choice than a physical one?

When safety, cost, access or repeatability make physical impractical.Always — simulations are better.Never — physical is always best.Only for older students.

8. A projectile simulation ignores air resistance. What should the teacher do?

State the assumption so students know the model is idealised.Say nothing; the simulation is exact.Tell students real projectiles behave identically.Avoid the simulation entirely.

9. Why does the prediction step matter in a prediction–observation–explanation activity?

It surfaces thinking; the surprise drives conceptual change.It does not matter.It lets students copy the answer.It replaces observation.

10. What is the difference between accuracy and precision?

Accuracy is closeness to the true value; precision is repeatability.They mean the same thing.Accuracy is repeatability.Precision is closeness to the true value.

11. A ruler is marked in millimetres. A spreadsheet reports a length as 12.4000 cm. What is wrong?

False precision — round to the significant figures the ruler supports.Nothing — more digits are always better.The unit is wrong.The ruler is too precise.

12. AI says two variables rising together means one causes the other. Why is this a problem?

Correlation is not proof of causation.Rising together always proves cause.It is a valid causal claim.Data can never mislead.

13. A line graph's y-axis starts at 20 instead of 0, making a small change look large. What is this?

A truncated axis that exaggerates the change.A correct, honest graph.A dual axis.A required design.

14. Why are control variables essential in a fair test?

They let you attribute the effect to the independent variable.They make the experiment longer.They are optional.They replace measurement.

15. What is the difference between a hazard and a risk?

A hazard can cause harm; risk is the likelihood and severity of harm.They are the same.A risk is a chemical.A hazard is always low risk.

16. In the control hierarchy, where does personal protective equipment (PPE) sit?

Last — after elimination, substitution, engineering and procedures.First — PPE is the main control.It is not part of the hierarchy.It replaces all other controls.

17. Is an AI-generated risk assessment sufficient to approve a laboratory activity?

No — the teacher makes the safety decision, never AI.Yes, if the AI seems confident.Yes, AI risk assessment is authoritative.Yes, for low-hazard activities only.

18. What is the difference between an error and a misconception?

An error is a slip; a misconception is a coherent wrong model.They are the same.An error is always a misconception.A misconception is easy to fix by stating the answer.

19. Why must you state the limit of every analogy?

Every analogy breaks down; the unstated limit can seed a misconception.Analogies have no limits.Limits confuse students unnecessarily.Analogies are always exact.

20. What must stay the same when you differentiate an explanation for diverse learners?

The core objective and scientific integrity.The reading level for everyone.Nothing needs to stay the same.The goal is lowered for some.

21. Why is starting a STEM project from a solution ('build a robot') a weak approach?

It skips the real need; scope the problem first.Solutions should always come first.Robots are always the best project.Problems do not matter.

22. How do prompt logs and AI-use disclosure support project integrity?

They keep AI use transparent and authorship clear.They hide how AI was used.They replace the students' work.They are unnecessary.

23. How should AI be used during project ideation?

To brainstorm and structure; students gather evidence and decide.To make the final decision for students.To replace all student thinking.It should not be used at all.

24. Who remains accountable for a decision made with AI support in science teaching?

The teacher remains accountable.The AI is accountable.No one is accountable.The tool's company only.

25. What is the difference between an AI prediction and a policy decision in environmental science?

A prediction is a model estimate; a policy decision is a human choice.They are the same.A prediction is always certain.A model makes the policy decision.

26. AI gives a claim citing a paper you cannot find, with an exact statistic. How should you treat it?

As unverified and likely fabricated; do not present it as fact.As true because AI is precise.As true if the statistic has decimals.Cite it directly to students.

27. What must a teacher check in an AI-drafted lab handout before use?

Accuracy, safety notes and accessibility.Only the font size.Nothing — AI handouts are reliable.Only the title.

28. What is 'answer leakage' in an assessment item?

A clue that reveals the answer without knowing the science.A correct answer.A hard question.A required feature.

29. A student's report includes a fabricated reference that AI invented. How should the teacher handle it?

As an integrity issue: fabricated references are never acceptable.Ignore it; AI references are fine.Accept it without disclosure.Only check the spelling.

30. A prompt for a science task should include which of these?

Context, role, grade, objective, materials, safety and output format.Just the topic.The students' real names.Nothing — AI needs no context.

31. Across the whole course, what is AI's role relative to the teacher?

A supervised assistant; the teacher verifies and stays accountable.A replacement for the teacher's expertise.The final scientific authority.A tool that makes safety decisions.

32. Which capstone criteria are pass-critical for certification?

Scientific accuracy, safety, privacy and teacher verification.Only how attractive the resources look.Only the word count.None — any package passes.

33. A generated image of a laboratory shows a student heating a flame with no eye protection. What should the teacher do?

Reject it; images must not show unsafe practices.Use it; images do not affect safety.Use it if it looks realistic.Add a caption and keep the unsafe image.

34. Should a downloadable resource marked 'Planned' be published to learners?

No — only real, working, verified resources.Yes — placeholders are fine.Yes — if clearly labelled.Yes — for premium courses.

35. During a graded assessment, what should the course AI assistant do?

Give hints and references, never the answers.Reveal the correct answers.Complete the assessment for the learner.Bypass the premium access check.