Exercises: Ethical Frameworks for the Data Age
These exercises progress from concept checks to challenging multi-framework analyses. Estimated completion time: 5-6 hours.
Difficulty Guide: - ⭐ Foundational (5-10 min each) - ⭐⭐ Intermediate (10-20 min each) - ⭐⭐⭐ Challenging (20-40 min each) - ⭐⭐⭐⭐ Advanced/Research (40+ min each)
Part A: Conceptual Understanding ⭐
Test your grasp of core concepts from Chapter 6.
A.1. Section 6.1.1 distinguishes between the legal and the ethical by offering three categories: "legal but unethical," "ethical but illegal," and "legal vacuum." In your own words, explain why each category exists — that is, why law and ethics do not perfectly overlap. Provide one original data-related example for each category that does not appear in the chapter.
A.2. Explain the difference between consequentialist and non-consequentialist ethical reasoning. Why does this distinction matter for data governance? Use utilitarianism and deontology as your primary examples, and identify one data governance question where the distinction leads to different conclusions.
A.3. Section 6.2.1 identifies four key features of utilitarianism: consequentialist, aggregative, impartial, and maximizing. For each feature, explain in one to two sentences why it is useful for data governance analysis, and in one to two additional sentences why it might be problematic. Use specific data-related examples where possible.
A.4. State Kant's categorical imperative in both formulations (universalizability and humanity as an end). Apply each formulation separately to the following practice: a social media company uses A/B testing to determine which notification frequency maximizes daily active users, without informing users they are subjects in an experiment. What does each formulation reveal?
A.5. Section 6.4.2 presents a table of five virtues for data practitioners. Choose any two virtues from the table and explain the "vice of excess" and "vice of deficiency" in your own words, using original examples drawn from data science, product design, or data governance. Why does Aristotle's concept of the mean matter for understanding ethical data practice?
A.6. Compare care ethics and deontology on the question of consent. How does each framework evaluate the practice of obtaining consent via terms-of-service agreements? What does each framework identify as the core ethical concern, and why do they differ?
A.7. Section 6.6.1 describes Rawls's veil of ignorance. Explain: (a) what information is hidden behind the veil, (b) why Rawls argues that rational people behind the veil would choose the difference principle, and (c) how the difference principle differs from a utilitarian approach to distributing the benefits and burdens of data systems. Use one concrete example from the chapter.
Part B: Applied Analysis ⭐⭐
Analyze scenarios, arguments, and real-world situations using the five frameworks.
B.1. Consider the following scenario:
A city government deploys a facial recognition system at public transit stations to identify individuals with outstanding warrants. The system has a 95% accuracy rate overall but a 78% accuracy rate for dark-skinned individuals. City officials argue that the system has led to 340 arrests in its first year and has made public transit safer, citing a 12% drop in reported assaults at stations.
Analyze this system from a utilitarian perspective and then from a justice theory (Rawlsian) perspective. Where do the two frameworks agree? Where do they disagree? Which framework do you find more persuasive in this case, and why?
B.2. Ray Zhao, in his guest lecture (Section 6.4.2), argues that the "hard cases" in data governance require judgment, not just rules. Using the virtue ethics framework, analyze the following situation:
A data scientist at a health insurance company discovers that a new risk model, while actuarially accurate, would significantly increase premiums for people in low-income zip codes — many of whom are racial minorities. Her manager says the model is "just following the data." She has not been asked to evaluate the model for fairness.
What virtues from the table in Section 6.4.2 are relevant? What would a person of practical wisdom (phronesis) do? How does the virtue ethics analysis differ from what a purely utilitarian analysis might recommend?
B.3. Section 6.5.2 applies care ethics to VitraMed's health data. Extend this analysis to the following scenario:
A children's hospital uses a predictive model to identify patients at risk of readmission within 30 days of discharge. The model requires sharing patient data (including diagnoses, family circumstances, and home addresses) with a third-party analytics company. The children's families signed consent forms, but the forms were presented in English only, and 40% of the hospital's patients come from Spanish-speaking families.
Analyze this scenario using care ethics. Who is in a relationship of vulnerability? What does "responsible care" require? How would a care ethics analysis differ from one based on deontological consent?
B.4. Apply the Rawlsian veil of ignorance to the following data policy:
An employer uses keystroke logging, email monitoring, and web browsing tracking for all employees. Employees were informed of the policy during onboarding. The company argues that monitoring improves productivity by 8% and helps detect potential data breaches. The data is stored for two years and is accessible to HR and management.
Behind the veil of ignorance, you do not know whether you will be the employer, a high-performing employee, an employee struggling with a personal crisis, an employee with a disability that affects typing speed, or an IT administrator with access to the monitoring data. What data governance policy would you design? Explain how the difference principle applies.
B.5. Section 6.3.2 distinguishes between using data to improve a service (treating users partly as means but also as ends) and harvesting behavioral surplus (treating users merely as means). Apply this distinction to the following three practices and determine, for each, whether it passes or fails the second formulation of the categorical imperative. Justify each answer:
- (a) A navigation app uses anonymized trip data to improve traffic predictions for all users.
- (b) A navigation app sells individual-level trip data to commercial real estate developers who use it to identify profitable locations for new stores.
- (c) A navigation app displays ads for nearby restaurants based on the user's current route, generating revenue that keeps the app free.
B.6. Mira says in Section 6.5.2: "Trust is a relationship, not a contract." Unpack this statement using care ethics and contrast it with a deontological analysis of trust. What does it mean for data governance if trust is relational rather than contractual? Identify one concrete governance practice that would follow from each view.
Part C: Multi-Framework Analysis ⭐⭐-⭐⭐⭐
These exercises require you to apply all five frameworks to a concrete dilemma and navigate their disagreements. Use the six-step ethical reasoning process from Section 6.7.1.
C.1. ⭐⭐⭐ The Genetic Database Dilemma. Read the following scenario and analyze it through all five frameworks using the six-step process.
GenomicsFirst, a direct-to-consumer genetic testing company, has a database of 4 million customers' DNA samples and associated health questionnaires. A government public health agency requests access to the de-identified database to study the genetic markers associated with a rare childhood disease that affects 1 in 50,000 children. The research could lead to early screening tests that save lives. However: (1) customers consented to "internal research" only, not government access; (2) genetic data is uniquely re-identifiable; (3) some customers belong to communities (indigenous groups, ethnic minorities) with historical reasons to distrust government use of biological data; and (4) the company would receive a $2 million research grant for providing access.
Step 1: Describe the situation. Step 2: Identify stakeholders (at least five). Step 3: Apply each of the five frameworks. Step 4: Identify convergences. Step 5: Identify divergences. Step 6: State your reasoned judgment and explain why you weight certain frameworks more heavily in this case.
C.2. ⭐⭐⭐ The Classroom Surveillance Debate. Analyze the following scenario through all five frameworks.
A public school district installs AI-powered cameras in every classroom that use facial expression analysis to monitor student engagement. The system flags students who appear distracted, confused, or disengaged, and sends alerts to teachers in real time. Administrators argue the system has improved test scores by 6% in pilot schools. Parents were notified via a letter sent home in backpacks. Teachers were not consulted before implementation. The technology vendor retains the data for "product improvement."
Apply the six-step ethical reasoning process. Pay particular attention to: (a) whether any framework provides a clear prohibition, (b) how care ethics handles the teacher-student relationship differently from the other frameworks, and (c) what the Rawlsian veil of ignorance reveals when you don't know whether you'll be a student, a parent, a teacher, an administrator, or a child with a neurodivergent condition whose facial expressions are systematically misread.
C.3. ⭐⭐⭐ The Pandemic Data Dilemma. Analyze the following scenario through all five frameworks.
During a global pandemic, a national government proposes a mandatory contact-tracing app. The app uses Bluetooth proximity data and GPS location to identify individuals who have been near confirmed cases. Refusing to install the app would result in a fine. Public health officials project that the app could reduce transmission by 30% and prevent an estimated 15,000 deaths over six months. Civil liberties organizations warn that the location data could be repurposed for immigration enforcement, political surveillance, or commercial use after the emergency ends. The app's source code is not open for public review.
Apply the six-step process. In your Step 6 judgment, explicitly address: Under what conditions, if any, can a public health emergency justify mandatory surveillance? What governance safeguards would change your analysis?
Part D: Synthesis & Critical Thinking ⭐⭐⭐
These questions require you to integrate, critique, and extend the frameworks presented in the chapter.
D.1. Section 6.7.2 argues for moral pluralism — the view that no single ethical framework captures everything that matters. But moral pluralism raises a challenge: if multiple frameworks are each partially right, how do you decide which framework to prioritize when they disagree? Using a specific data governance dilemma (you may use one from the chapter or create your own), develop a set of criteria for when each framework should be given priority. For example: "When vulnerable populations are disproportionately affected, care ethics and justice theory should be weighted more heavily than utilitarian calculations." Propose at least three such criteria and defend each one.
D.2. The chapter presents five ethical frameworks but does not discuss several others that scholars have applied to data governance, including: - Contractualism (T.M. Scanlon — actions are wrong if they violate principles that no one could reasonably reject) - Ubuntu (African communitarian philosophy — "I am because we are") - Confucian ethics (role-based obligations, relational harmony)
Choose one of these alternative frameworks (or another you have encountered in prior coursework). In 400-500 words, explain: (a) its core principles, (b) what it would say about a data governance dilemma discussed in the chapter, and (c) what it reveals that the five frameworks in the chapter do not. Should the chapter have included it? Why or why not?
D.3. Eli argues in Section 6.3.2 that Kant would reject surveilling a community without consent regardless of whether surveillance makes the neighborhood safer. Mira pushes back by asking whether the consequences matter. Dr. Adeyemi sides with the Kantian view.
Construct the strongest possible counter-argument to the Kantian position — that is, the best case for a consequentialist approach to community surveillance. Then construct the strongest possible counter-argument to the consequentialist position. Which argument do you find more persuasive, and what does your answer reveal about your own ethical intuitions?
D.4. The virtue ethics section (6.4) focuses on individual character — what a good data practitioner should be. But many ethical failures in data governance are structural, not individual — they arise from business models, incentive systems, regulatory gaps, and organizational cultures. Is virtue ethics equipped to address structural problems? Or does it risk placing too much responsibility on individuals while leaving harmful structures intact? Write a two-paragraph analysis that engages with this tension, drawing on at least two examples from the chapter.
D.5. Consider the following claim:
"Rawls's veil of ignorance is impractical because real decision-makers always know their position in society. A CEO cannot genuinely pretend not to know she is a CEO. The thought experiment is philosophically interesting but useless for actual data governance."
Evaluate this critique. Is the veil of ignorance useful despite its impracticality? Can you identify governance mechanisms that attempt to approximate the veil's effects (i.e., structural arrangements that force decision-makers to consider perspectives other than their own)? Provide at least two examples.
Part E: Research & Extension ⭐⭐⭐⭐
These are open-ended projects for students seeking deeper engagement. Each requires independent research beyond the textbook.
E.1. Shannon Vallor's "Technology and the Virtues" — A Research Essay. Section 6.4 presents a general overview of virtue ethics applied to data. Shannon Vallor's Technology and the Virtues: A Philosophical Guide to a Future Worth Wanting (2016) develops a much more detailed framework. Read Chapters 1, 4, and 7-8 of Vallor's book and write a 1,500-word essay that addresses: (a) How does Vallor adapt Aristotelian virtue ethics for the technological age? (b) What "technomoral virtues" does she identify, and how do they compare to the virtues listed in Section 6.4.2? (c) What does Vallor's framework add to the analysis of data governance that the chapter's treatment does not cover? Use at least three sources beyond this textbook.
E.2. The Ethics of Predictive Policing: A Five-Framework Case Study. Research a specific predictive policing system (PredPol/Geolitica, HunchLab, the Chicago Strategic Subject List, or another system you identify). Write a 2,000-word case study that: (a) describes the system, its data sources, and its deployment context, (b) analyzes the system through all five ethical frameworks from the chapter, (c) identifies where frameworks converge and diverge, and (d) provides your reasoned judgment on whether the system should be deployed, modified, or discontinued. Use at least five sources, including at least one source from a community affected by the system.
E.3. Care Ethics and AI: An Emerging Field. Care ethics has traditionally focused on interpersonal relationships, but a growing body of scholarship applies it to technology and AI systems. Research this emerging field by reading at least two of the following: Virginia Held's The Ethics of Care (2006), Mark Coeckelbergh's "Artificial Companions: Empathy and Vulnerability in the Age of Artificial Intelligence" (2010), or Jathan Sadowski's work on data as care infrastructure. Write a 1,200-word essay that: (a) identifies the key challenges of scaling care ethics to data systems, (b) proposes at least two specific governance practices that operationalize care ethics principles, and (c) evaluates whether care ethics offers something that rights-based (deontological) frameworks cannot.
E.4. Ethical Framework Debate: A Structured Exercise. Organize a structured debate (solo or with classmates) on the following proposition:
"Resolved: In cases of conflict between ethical frameworks, utilitarian analysis should be the default framework for data governance decisions because it alone provides a systematic method for comparing outcomes."
Prepare a 1,000-word brief arguing for the proposition and a 1,000-word brief arguing against it. Each brief should: (a) articulate the strongest version of its position, (b) anticipate and rebut the strongest counterargument, and (c) use at least two real-world data governance cases as evidence. Conclude with a 300-word personal reflection on which side you found more persuasive and why.
E.5. Moral Pluralism in Practice: Institutional Case Study. Research an organization that has attempted to operationalize multiple ethical frameworks in its technology ethics processes. Possibilities include: Google's AI Principles and their implementation challenges, Microsoft's Responsible AI program, the IEEE's Ethically Aligned Design framework, or a university's IRB process for data research. Write a 1,500-word analysis that: (a) describes the organization's ethical framework(s), (b) evaluates whether the framework reflects moral pluralism or privileges one ethical tradition, (c) identifies where the framework succeeds and where it falls short, and (d) proposes one specific improvement. Use at least four sources.
Solutions
Selected solutions are available in appendices/answers-to-selected.md.