Writing Good Questions for Virtual Labs (10 Examples)

Good questions for virtual labs make students use the simulation, not their memory, and every wrong answer tells you something. This guide is about the craft: how to word a question, which type to choose, how to set tolerances and accepted answers, how to write distractors that diagnose misconceptions, and how to word a prediction. It ends with ten ready-to-copy examples across subjects, each with the settings to pin, the exact answer and a weak version to avoid. Setting up a question set is covered in collecting student answers without accounts.

Five rules for wording

1. Name the setting and the readout. "Set V to 1.2 L. What does the gauge read?" beats "What is the pressure?" Students working alone can't ask which value you meant.
2. Know what the screen prints. If the answer is printed on the screen, the question only checks that students found the right readout. That is fine for a first reading. To make students think, ask for a comparison, a threshold, an input that gives a target, or a reason.
3. Put the unit and format in the question. "In kilometers", "as a percentage". The answer box shows your unit, but students still type what they think you asked for.
4. One idea per question. When a two-part question is wrong, you can't tell which part failed.
5. Pin everything random. If the simulation has random parts, pin them off, or ask about something that doesn't depend on chance.

Choose the type by what you need to know

You want to know…	Use	Grading
Can they read or calculate a value?	Number, ± value	Answer and tolerance
Can they calculate through several steps?	Number, ± %	Tolerance grows with the answer
Can they find any setting that works?	Number, range	Minimum and/or maximum, each "included" or not
Is there more than one right number?	Number, list of values	Up to 10 accepted values
Which idea do they hold?	Multiple choice	2–5 options, one correct
Can they name something?	Short answer	Up to 10 accepted answers, * as a wildcard
Can they explain?	Short answer	No accepted answers; read them yourself

Numbers are forgiving about format. Students can type a comma or a point for decimals, and they can add the unit at the end. Short answers ignore capital letters and extra spaces unless you tick Case sensitive.

Tolerances: cover honest error, nothing more

• Readings: match the display. If the simulation shows two decimals, ± 1 in the last digit is enough.
• Calculated values: use a percentage, because rounding errors grow with the size of the answer.
• Counts and positions: tolerance 0.
• Random results: use a range wide enough for honest runs.

Run the experiment yourself first. If the simulation rounds and students calculate, check that both answers pass.

Distractors that diagnose

Write each wrong option from a real misconception, so the results table tells you which idea students hold. An option nobody believes, such as "ten times faster", wastes a slot. With number questions, choose values so that each common error produces a different wrong number. Then a cluster of identical wrong answers in the results points to one mistake.

Wording predictions

Prediction questions are asked before the simulation unlocks, and only multiple choice and number questions can be asked again afterward. Word predictions so they need a mental model, not a fact: "What will happen to… if…?" Give the starting situation in the question, because students can't see the simulation yet. Predictions are never scored. Tell students, so they write what they really think.

Ten examples to copy

Each example is one question on its own simulation, with values taken from the simulation's model. Add it to a set with up to four other questions on the same simulation.

1. Biology: heterozygotes (number, ± value)

Hardy–Weinberg equilibrium. Pin: p = 0.6, selection 0, genetic drift off.

• Question: "What percentage of the population is heterozygous (Aa)? Give a percentage."
• Answer: 48, ± 0.5, unit %.
• Why it works: with drift off, every student sees the same bar, Aa 48.0%. The format is in the question.
• Weak version: the same question with drift on. The value wanders every generation, so correct students get marked wrong.

2. Geography: map distance (number, ± %)

Map scale and geographic coordinates. Pin: scale 1:25 000.

• Question: "Press Reset A, B. What ground distance does AB represent, in kilometers?"
• Answer: 2.1, ± 3%, unit km.
• Why it works: the simulation shows 8.5 cm and rounds the ground distance to 2.1 km. Students who calculate get 8.5 × 250 m = 2.125 km. A 3% tolerance accepts both.
• Weak version: "How far is it from A to B?" Some answer in centimeters, some in meters.

3. Physics: a threshold (number, range)

Boyle's law. Pin: initial volume 3 L, initial pressure 1 atm.

• Question: "Find a volume at which the gauge reads more than 4 atm. Type the volume in liters."
• Answer: range, minimum 0.5 (included), maximum 0.75 (not included), unit L.
• Why it works: many right answers, and it tests the inverse relationship. At 0.75 L the gauge reads exactly 4.00 atm, so 0.75 is excluded. Students who think pressure rises with volume type a large volume and are marked wrong.

4. Computer science: more than one answer (number, list of values)

Linear search vs. binary search. Pin: 20 elements, Both (compare), array sorted.

• Question: "Click a cell to set the target, then press Play. Give one position (1–20) where linear search needs fewer comparisons than binary search."
• Answer: list of values 1, 2, 3, 4.
• Why it works: the array values are random, but the comparison counts depend only on the position. Binary search needs at most 5 comparisons for 20 elements, so linear search only wins in the first four positions.
• Weak version: "How many comparisons did binary search take?" It depends on the target each student clicked.

5. Math: base rates (multiple choice prediction, asked again)

Conditional probability and Bayes' theorem. Pin: prevalence 1%, sensitivity 95%, specificity 90%.

• Question: "1% of people have a disease. A test is 95% sensitive and 90% specific. Mia tests positive. What is the chance she has the disease?"
• Options: about 95% / about 90% / about 50% / about 9%.
• Answer: about 9%. The simulation shows 8.76%: 95 true positives among 1,085 positive tests out of 10,000 people.
• Why it works: each distractor is a known error. 95% confuses sensitivity with P(disease | positive), 90% uses the specificity, 50% is "it's either yes or no".

6. Physics: wrong answers that diagnose (number)

Ohm's law – two resistors in series. Pin: U = 12 V, R1 = 10 Ω, R2 = 20 Ω, show readings on.

• Question: "In the starting values under the simulation, change R2 to 40 Ω. What is the voltage across R1?"
• Answer: 2.4, ± 0.05, unit V.
• Why it works: 4.0 means "R1 didn't change, so its voltage didn't either". 6.0 means "the voltage splits equally". 9.6 means the student read U2. Each error has its own number.
• Explanation to add: "The current falls from 0.4 A to 0.24 A, so U1 = 0.24 × 10 = 2.4 V."

7. Chemistry: naming (short answer, accepted answers)

Halide ions – identification with AgNO₃. Pin: NaCl.

• Question: "Test all four solutions. Which halide ion gives a precipitate that dissolves completely in excess ammonia? Give its name or symbol."
• Accepted answers: *chloride*, Cl-, Cl⁻, Cl.
• Why it works: the list covers the name, the typed symbol and the superscript symbol. "Chlorine" is not accepted. It is a real error, so read those answers.
• Weak wildcard: *cl* also accepts "nuclear" and "include". Keep wildcards around whole words.

8. Chemistry: a prediction that surprises (number, asked again)

Calorimetry – enthalpy of neutralization. Pin: 50 mL of 1.0 M HCl, 50 mL of 1.0 M NaOH, heat loss 5%.

• Question: "With 50 mL of each solution, the temperature rises by about 6.4 °C. Predict the rise with 100 mL of each, at the same concentrations."
• Answer: 6.4, ± 0.3, unit °C.
• Why it works: a common prediction is 12.8 °C. Twice the volume releases twice the heat but also has twice the mass to warm. The starting result is in the question, because the simulation is locked during predictions.

9. Biology: using evidence (short answer, read by hand)

Bacteria in food – the temperature danger zone. Pin: Bacillus cereus, 25 °C, 10 hours, reheat on.

• Question: "Your cousin says rice left out overnight is safe if you reheat it. Reply in two sentences and quote one number from the simulation."
• Model answer: "After 10 hours at 25 °C the count reaches about 10^{7.9 CFU/g, far above the illness level of 10}5. Reheating kills the bacteria, but the toxin they already made survives, so the rice is still unsafe."
• Why it works: it needs a claim and evidence, and the context is real. Leave the accepted answers empty and read these yourself.

10. Physics: two-tier (two multiple-choice questions)

Centripetal force. Pin: m = 0.5 kg, R = 1 m, ω = 3 rad/s, vectors on.

• Question A (prediction, asked again): "The string is cut. The mass moves…" Options: along the tangent / straight outward along the radius / in a spiral outward / in a circle for a while. Answer: along the tangent.
• Question B: "Why?" Options: no inward force acts, so it keeps its velocity / a centrifugal force throws it outward / it still has circular momentum. Answer: no inward force acts.
• Why it works: students can guess the path; they can't guess the reason as well. Tier B separates lucky guesses from understanding.

For using results to adjust your teaching, see formative assessment with simulations. For more classroom ideas, start with how to use interactive simulations in the classroom.

FAQ

How many questions should a virtual lab have?

A set holds up to five. Two or three numbers, one multiple choice and one short answer is a good mix for a 30-minute lab.

Should I allow retries?

Yes, for readings. One or two retries send students back to the simulation to re-measure. Retries need feedback that shows right or wrong.

What is the best tolerance for a number question?

Run the lab yourself. For readings, cover the last displayed digit. For calculated results, use 2–5%. For counts, use 0.

Can short answers be graded automatically?

Yes, if the answer is a word or a short phrase. Add up to 10 accepted answers. For explanations, leave the list empty and read them.