Skip to content

MODULE 10 · LESSON 56 · Advanced application

Data ethics and privacy

Minimize sensitive information and consider how collection, grouping and recommendations affect people.

By Laurence Svekis · Self-paced · All worked data is synthetic unless you supply your own source.

Try the practice ↓

PredictInspectCalculateCheckExplainTransfer

Your starting point

Prerequisites: Use the ideas and vocabulary from lesson 55; revisit its worked examples if needed.

Definition: Data ethics concerns responsible choices about collecting, using, interpreting and sharing information. Privacy concerns appropriate access and control over information about people. De-identification removes or changes identifiers, but does not automatically eliminate re-identification or harmful inference risks.

Use only what the question needs

A useful analysis often needs fewer fields than an available export contains. Remove unnecessary names, contact details and granular histories before creating practice materials. Prefer synthetic examples for learning. Check the data owner’s permissions and applicable organizational rules before handling real records or using external AI services. The academy cannot determine whether a particular use is authorized.

Consider combinations and small groups

A record can be recognizable through a rare combination of location, date and role even after names are removed. Aggregates can also expose information when a group has very few members. Avoid publishing details that allow individuals to be inferred without a legitimate, reviewed purpose. There is no universal small-cell threshold that makes every context safe; assess the actual information and audience.

Look beyond confidentiality

An analysis can be technically private yet unfairly interpreted. Group averages should not become unsupported judgments about individuals. Missing groups and biased measurements can lead to recommendations that neglect people not represented in the data. Document whose interests are affected, how limitations constrain the decision and who can challenge an error. Responsible practice includes transparency and proportionality, not merely hiding a name column.

Two worked examples

A recognizable record

A fictional table removes names but retains exact appointment date, a tiny workplace and a rare job role. Someone familiar with the workplace might still recognize the person. Using a synthetic example or a less granular aggregate may better fit a training purpose without exposing the original combination.

An unfair inference

A dashboard shows a lower average outcome for one cohort. It does not establish that every member has lower capability or that cohort identity caused the difference. Using the aggregate to label an individual would ignore within-group variation, confounding and the limits of the measurement.

Where this becomes useful

An educator preparing a workshop can use the academy’s synthetic datasets instead of real learner records. The same analytical skills can be practised without introducing avoidable exposure of personal information.

Make an attempt

Open a synthetic dataset in the workbench and draft a privacy note for a hypothetical real version. Identify unnecessary fields, potentially identifying combinations, intended audience and a review step before any external sharing.

Open the accompanying interactive lab →

Misconception check

Claim: Keeping a dataset local resolves every ethical question.

Inspect the correction

Local processing limits transmission by this tool, but authorization, interpretation, retention and onward sharing still require judgment.

Independent application

Write an ethical-use note for a fictional analysis. State the purpose, minimum data needed, access and retention plan, possible affected groups, interpretation limits and a route for reviewing errors. Use synthetic records for the demonstration.

Completion records practice and self-review. It is not certification of statistical competence.