CDEF study plan

Define and contrast the major normative ethical theories (consequentialism, deontology, virtue ethics, care ethics, contractualism) in your own written words, listing the decision rule each theory supplies for a data-handling dilemma.

Map each ethical theory to at least two concrete data-practice scenarios (e.g., re-identifying a dataset, selling inferred profiles) and explain which theory condemns or permits the act and why.

Distinguish descriptive ethics, normative ethics, and applied ethics, and classify ten data-ethics claims you encounter in news coverage into those categories.

Explain the concepts of informed consent, moral residue, and conflicting duties as they arise when ethical obligations collide with business goals.

Identify the ethical values underlying privacy (dignity, autonomy, contextual integrity) and explain how each differs from the purely legal notion of compliance.

Apply Nissenbaum's contextual integrity framework to three data flows and judge which violate contextual norms of appropriateness or distribution.

Distinguish the categories of personal data, sensitive data, pseudonymized data, and anonymized data, and explain why anonymization claims require scrutiny.

Analyze how secondary use, data brokerage, and purpose limitation interact, and argue for or against a specific secondary-use practice.

Explain the data-subject rights typically granted by modern data-protection regimes (access, rectification, erasure, portability, objection) and describe one practical obstacle to each.

Evaluate the ethical limits of consent as a legitimizing basis, including consent fatigue, dark patterns, and the problem of group harms that individuals cannot consent away.

Compare individual-rights and collective/group privacy approaches, and argue when group-level protections are ethically required.

Assess trade-offs between privacy and utility in a specific case (e.g., contact tracing, fraud detection) and justify a proportionate design.

Distinguish the main fairness definitions (demographic parity, equalized odds, predictive parity, calibration, individual fairness) and state the conditions under which each is appropriate.

Demonstrate the mathematical impossibility of satisfying several fairness criteria simultaneously under unequal base rates, and explain the practical implication for model selection.

Trace how bias enters each stage of the data lifecycle (problem formulation, collection, labeling, feature engineering, modeling, deployment) and name a mitigation for each stage.

Identify proxy variables and historical bias in a dataset description, and justify which variables you would remove or reweight.

Explain the difference between transparency, interpretability, and explainability, and match each to the stakeholder who needs it (affected individual, regulator, developer, auditor).

Compare inherently interpretable models with post-hoc explanation methods, and state the fidelity and stability risks of post-hoc explanations.

Analyze the right to explanation as an ethical and legal demand, including when explanations are owed and what content makes one meaningful.

Assign accountability across roles (data owner, model developer, deployer, auditor) for a harmful automated decision, and identify where accountability gaps arise.

Draft the core components of a data-governance framework: roles, data classification, retention, access control, provenance, and audit trails, and explain the ethical rationale for each.

Compare the legal instruments most relevant to data ethics (GDPR-style data-protection law, sectoral rules, AI-specific regulation, anti-discrimination law) by scope, obligations, and enforcement.

Explain how legal compliance and ethical adequacy can diverge, using at least two examples where lawful conduct remains ethically contested.

Evaluate security-ethics dilemmas such as breach disclosure timing, encryption backdoors, and re-identification research, and justify a defensible position.

Apply at least two structured decision frameworks (e.g., a principles-based test, a stakeholder-impact analysis, a precautionary approach) to the same case and compare the conclusions.

Construct a full ethical case analysis: facts, stakeholders, values, options, harms, trade-offs, decision, justification, and monitoring plan.

Identify common reasoning failures in data-ethics decisions (technological solutionism, diffusion of responsibility, metric fixation, hindsight bias) and describe a countermeasure for each.

Evaluate a documented AI or data-ethics failure and attribute the root causes to specific decision points rather than to individuals alone.

Assess the societal-impact questions raised by generative AI, large-scale model training data, biometric surveillance, and synthetic media, and state the ethical stakes of each.

Explain how organizational incentives, metrics, and reporting lines shape ethical conduct, and propose two structural changes that reduce misconduct risk.

Describe the professional obligations of a data practitioner who discovers harmful practice, including escalation paths, documentation, and the ethics of whistleblowing.

Consolidate the whole syllabus by re-testing your weakest areas and correcting the misconceptions your practice results reveal.