Professional and Ethical Considerations in Data Science
What the exam tests
This domain tests professional judgement when data use creates ethical, legal and reputational risk.
Core rules
- The IFoA Actuaries' Code requires integrity, competence and care, impartiality, compliance and clear communication. Data work is actuarial work when it informs decisions.
- GDPR principles: lawful basis, purpose limitation, data minimisation, accuracy, storage limitation, integrity and confidentiality, and accountability.
- Fairness: test for disparate impact, consider proxies, and document trade-offs between accuracy and fairness.
- Explainability and accountability: keep a human accountable, provide meaningful explanations, and allow challenge.
- Third-party data and outsourcing remain your responsibility; document due diligence.
Worked example
An actuarial team builds a fraud model using device data. The model flags customers with unusual app usage. The team finds the flag correlates with disability status through accessibility settings. They remove the device-accessibility variables, retest performance, document the residual risk, and add human review before any adverse action. They record the decision in the model documentation and inform the model risk committee. The reasoning is that legal risk and fairness outweigh a small AUC gain, and human review preserves a route to challenge.
Traps
- Treating anonymisation as automatically sufficient; re-identification is possible.
- Assuming the data team owns the ethics decision; accountability sits with actuarial sign-off.
- Confusing legal compliance with ethical acceptability.
- Failing to document why a variable was excluded.