016 min
One clue, many matches
Read a broad fictional age clue and mark all matches on paper.
Ask: “Does this clue name one person?”
Listen for: “No, lots of records fit.”
Brightlab
brightlab.Made for curious mindsbrightlab.Made for curious mindsBrightlab
Year 2 · Privacy
One clue fits many make-believe people. What happens when clues join?
Go to the investigation ↓A minimal survey design with candidate counts and a statement of residual risk.
One clue fits many make-believe people. What happens when clues join?
One broad clue can match many fictional people. Several precise clues together may match only one. Our mosaic counts how many supplied synthetic records fit an age range and map area. A larger matching group can reduce this kind of identification risk, but it is not a guarantee of privacy: another clue could narrow the group again.
Before this lesson: Group fictional cards by a visible feature. Useful earlier investigations: A message with just enough.
One broad clue can match many fictional people. Several precise clues together may match only one. Our mosaic counts how many supplied synthetic records fit an age range and map area. A larger matching group can reduce this kind of identification risk, but it is not a guarantee of privacy: another clue could narrow the group again.
Group fictional cards by a visible feature. Useful earlier investigations: A message with just enough.
Print the supplied fictional grid of 24 records. Explain clearly that no record describes a class member. Set the survey purpose: count people in a broad region and age band, not find one person.
ACARA V9 AI curriculum connection ↗ · Technologies ↗
A complete teaching sequence · 45 minutes
Read a broad fictional age clue and mark all matches on paper.
Ask: “Does this clue name one person?”
Listen for: “No, lots of records fit.”
Add a location clue and predict whether the matching crowd grows or shrinks.
Ask: “Can adding another clue create more matches?”
Listen for: “It should keep or remove matches.”
Narrow only the age range, then reset and narrow only the map region. Record remaining candidates and retained regional usefulness.
Ask: “Which records disappeared, and why?”
Listen for: “They no longer fit the smaller window.”
Apply both precise fields and reveal a unique synthetic match.
Ask: “Were the clues still harmless when combined?”
Listen for: “Together they pointed to one record.”
Coarsen fields until several records match while the broad-region count remains possible. Write a purpose statement.
Ask: “What detail does our question actually need?”
Listen for: “An age band and region, not an exact age and address.”
Explain why a count of several matches is not a promise of anonymity.
Ask: “Could a new clue narrow it again?”
Listen for: “Yes, like another known detail.”
Harmless facts cannot identify someone.
Age band and location each match multiple fictional records, but their intersection can match one.
Choose coarse age and region fields for a regional survey and record both usefulness and remaining matches.
Ask learners to describe an intersection using the highlighted records. Do not accept a fixed candidate count as a universal privacy guarantee.
| Criterion | Beginning | Secure | Extending |
|---|---|---|---|
| Combining clues | Considers each field alone | Explains that combining narrows matches | Predicts and verifies an intersection |
| Minimisation | Deletes data without considering purpose | Coarsens fields while preserving purpose | Explains residual identification risk |
Use eight fictional tokens and two circles to show overlap physically before exploring numbers.
Add a third invented clue on paper and explain why a previous privacy judgement needs revision.
Never enter real class demographics, addresses or birthdays. This demonstration is not a method for identifying real people.
Shared device? Turn remembering off. A project file lets you continue on another device.
Your browser is the laboratory
One clue fits many make-believe people. What happens when clues join?
Twenty-four fictional records cover ages 8–13 in each of four regions. The age clue begins at 10 and can include one to four ages; the region clue can include one to four regions.
The experiment opens after your prediction.
There is no penalty for being surprised.
Apple Silicon · PyTorch MPS
Generate large synthetic populations and compare uniqueness rates under different age/location coarsening policies, keeping the underlying population fixed.
Teacher-operated extension: students predict and interpret the visual report together.
The bundle contains lesson-specific working code, a configuration file, a reactive notebook, a deterministic CPU check and hardware setup instructions. Acceleration is reported only after a tensor operation and result read-back succeed.
Download Mac Studio investigation ↓unzip y2-mosaic-mac-pathway.zip -d y2-mosaic-mac cd y2-mosaic-mac bash setup-mac.sh source .venv/bin/activate python experiment.py --device mps --output results marimo edit notebook.py
CPU and available-device execution status is recorded in the downloaded README and validation report. DGX Spark execution requires that hardware; static validation alone does not establish GPU compatibility or performance. The browser lesson remains fully available without this extension.
NVIDIA DGX Spark · PyTorch CUDA
Generate large synthetic populations and compare uniqueness rates under different age/location coarsening policies, keeping the underlying population fixed.
Teacher-operated extension: students predict and interpret the visual report together.
The bundle contains lesson-specific working code, a configuration file, a reactive notebook, a deterministic CPU check and hardware setup instructions. Acceleration is reported only after a tensor operation and result read-back succeed.
Download DGX Spark investigation ↓unzip y2-mosaic-dgx-pathway.zip -d y2-mosaic-dgx cd y2-mosaic-dgx bash run-dgx.sh # Open the localhost notebook URL printed by the container.
CPU and available-device execution status is recorded in the downloaded README and validation report. DGX Spark execution requires that hardware; static validation alone does not establish GPU compatibility or performance. The browser lesson remains fully available without this extension.
A minimal survey design with candidate counts and a statement of residual risk.
Download the editable handout →