Foundations of Data Ethics and Ethical Theory
Scope
Foundations tests your ability to name and apply ethical theory to data practice, and to separate ethical reasoning from legal compliance. Expect items that ask which framework best justifies a decision, or that ask you to spot an argument that is really a compliance argument in disguise.
Core frameworks
- Consequentialism / utilitarianism - judges acts by outcomes and maximises aggregate welfare. It forces you to count harms and benefits, but it can license sacrificing a minority for aggregate gain, and it suffers measurement problems when harms are diffuse.
- Deontology - duties and rights. Kant's formula of humanity forbids treating persons merely as means; this maps directly onto consent, dignity and anti-instrumentalisation arguments.
- Virtue ethics - asks what a careful, honest, just practitioner would do. Strong on character and professional judgement, weak at resolving conflicts between duties.
- Justice theories - Rawls' difference principle holds that inequalities are justified only if they benefit the least advantaged. This is the standard lens for access and allocation questions; capabilities approaches extend it to what people can actually do.
- Principlism - Beauchamp and Childress's four principles (autonomy, beneficence, non-maleficence, justice) dominate applied ethics. It is a checklist, not an algorithm: principles conflict and the balance must be justified in reasons.
- Care ethics - foregrounds relationships, dependency and the concrete other; useful for health and social-service data.
Also know the Belmont Report's three principles (respect for persons, beneficence, justice) as the ancestor of modern research review, and information ethics (Moor, Floridi) as broadening moral concern beyond human agents to the informational environment.
Worked example
A hospital deploys a sepsis model trained on five years of admissions. It under-performs for uninsured patients. A utilitarian argues net lives saved justify deployment. A deontologist objects that patients are being used as means to population benefit without their knowledge. A Rawlsian asks whether the error-rate inequality benefits the least advantaged - it does not. Principlism flags justice and non-maleficence. The defensible answer combines them: conditional deployment with subgroup monitoring, governance oversight and mitigation, because the frameworks converge on the subgroup harm even where they differ on why it matters.
Traps
Equating lawful with ethical - law is a floor and often lags. Assuming consent cures all harms, including group and third-party harms. Treating one framework as decisive when the question asks you to compare. Forgetting that the same act can be justified by different theories - items often want the theory, not the conclusion.