Exercises: Building a Data Ethics Program

These exercises progress from concept checks to challenging applications. Estimated completion time: 3-4 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 26.

A.1. Section 26.1.1 presents the "compliance floor" concept and Ray Zhao's anecdote about a VP calculating "acceptable risk." Explain in your own words why a cost-benefit analysis of regulatory fines is insufficient as a basis for data ethics. What does the VP's reasoning miss?

A.2. Define ethics-washing and explain why it is potentially worse than having no ethics program at all. Use the concept of legitimacy in your answer.

A.3. Section 26.2.2 identifies six composition elements essential for an ethics committee. For each, explain in one sentence what blind spot would result from the element's absence.

A.4. Explain the difference between an advisory ethics board and a gate-keeping ethics board (Section 26.2.3). Why does Ray Zhao describe the shift to advisory-with-escalation as transformative at NovaCorp?

A.5. Section 26.3.1 acknowledges that philosophical frameworks from Chapter 6 are "operationally abstract." In your own words, explain why an engineer with a launch deadline cannot simply "apply utilitarianism" to a product decision without additional tools.

A.6. What does "tone at the top" mean in the context of data ethics programs? Why is leadership commitment a necessary — but not sufficient — condition for ethical culture?

A.7. Section 26.4 discusses incentive alignment. Explain the concept of "perverse incentives" and provide an example (not from the chapter) of how a performance metric could discourage ethical data practices.

Part B: Applied Analysis **

Analyze scenarios, arguments, and real-world situations using concepts from Chapter 26.

B.1. Consider the following scenario:

A mid-size social media company announces a "Responsible Data Committee" composed of the CEO, CTO, General Counsel, VP of Product, and VP of Engineering. The committee meets quarterly to review "any ethical concerns raised by employees." It publishes a one-page annual report on its activities. The committee's recommendations are advisory; the CEO makes all final decisions.

Using the criteria from Section 26.2, evaluate this committee's design. Identify at least four specific weaknesses and, for each, propose a concrete improvement.

B.2. Section 26.3.2 introduces the ethical risk assessment matrix. Apply this matrix to the following proposal:

VitraMed is considering selling aggregate, de-identified patient wellness scores to an employer wellness program that wants to compare the health metrics of its workforce against industry benchmarks. The data would be aggregated at the company level (minimum group size: 50 employees) with no individual-level data shared.

Identify at least three ethical risks, rate each for likelihood and severity, and propose mitigations.

B.3. Ray Zhao describes the CDO role as "being the person who's responsible for the roof but doesn't own any of the walls" (Section 27.1.3, referenced in class discussion). Apply this metaphor to the role of an ethics program leader. What "walls" must an ethics program navigate, and what happens when the "wall owners" resist?

B.4. Section 26.5 discusses culture change strategies. Analyze the following approach:

A company's new Chief Ethics Officer emails all 3,000 employees a 15-page "Data Ethics Handbook" and requires them to pass a 50-question online quiz within 30 days. Employees who pass receive a "Data Ethics Certified" badge for their email signature.

Evaluate this approach against the culture change principles from Section 26.5. What is likely to work? What is likely to fail? What would you do differently?

B.5. Mira's exchange with her father in Section 26.1.2 raises a tension between business viability and ethical practice. Vikram says, "I'm running a business." Construct an argument that responds to Vikram without dismissing his concern — one that demonstrates how ethics and business viability can be aligned. Then identify at least one scenario where they genuinely cannot be aligned, and discuss what should happen in that case.

B.6. Section 26.6 warns about the dynamics that produce ethics-washing. For each of the following organizational behaviors, explain whether it constitutes ethics-washing, genuine ethics practice, or something in between. Justify your classification.

• (a) A company publishes AI ethics principles on its website but has no internal process for reviewing AI products against those principles.
• (b) A company hires a Chief Ethics Officer who reports to the General Counsel and has a team of three, but gives the role veto power over high-risk product launches.
• (c) A company donates $5 million to a university AI ethics research center and cites this donation in response to media inquiries about its own data practices.
• (d) A company conducts quarterly ethics reviews of all new data collection practices but never publishes the results externally.

Part C: Real-World Application Challenges -*

These exercises ask you to investigate real-world ethics programs and apply the chapter's frameworks.

C.1. ** Corporate Ethics Audit. Select a major technology company (e.g., Microsoft, Google, Meta, Amazon, Apple, or Salesforce). Research its publicly available responsible AI or data ethics program. Using the framework from Section 26.2, evaluate:

• (a) What ethics governance structures exist (committees, boards, officers)?
• (b) What authority do these structures have (advisory, gate-keeping, veto)?
• (c) What independence mechanisms are in place?
• (d) What transparency exists about the program's activities and outcomes?

Write a one-page assessment rating the program as decorative, advisory, advisory-with-escalation, gate-keeping, or veto-capable. Support your rating with evidence.

C.2. ** Ethics Decision Tree. Section 26.3.3 introduces the ethical decision tree. Design a decision tree for the following context: a university's Office of Institutional Research is considering whether to use student learning management system (LMS) data (login times, page views, assignment submission patterns) to build a predictive model for student dropout risk. Your decision tree should include at least six decision points, each grounded in a specific ethical consideration from the chapter.

C.3. *** Incentive Design Exercise. You are the newly appointed ethics program director at a company with 500 data scientists and engineers. The current incentive structure rewards speed of deployment and user growth metrics. Design a revised incentive structure that incorporates ethical data practice without undermining productivity. Your proposal should include:

• At least three specific ethical performance metrics
• A method for measuring each metric
• An explanation of how these metrics integrate with existing performance evaluation
• A response to the objection: "This will slow us down"

C.4. *** Stakeholder Mapping. Choose a data-intensive product you use regularly (a recommendation engine, a ride-sharing app, a fitness tracker, a financial app). Map the stakeholders affected by this product's data practices. For each stakeholder group, identify: (a) what ethical obligations the company has toward them, (b) whether those obligations are currently being met (based on publicly available information), and (c) what an ethics committee reviewing this product should flag.

Part D: Synthesis & Critical Thinking ***

These questions require you to integrate multiple concepts and think beyond the material presented.

D.1. The chapter presents a tension between two failure modes: ethics programs that are too weak (decorative, advisory-only) and ethics programs that are too strong (so burdensome that teams circumvent them). Drawing on the principles of proportionality, threshold testing, and culture change discussed in this chapter and Chapter 28, propose a framework for calibrating the "strength" of an ethics program to the level of risk. How should an organization decide which data practices need full ethics review and which need only lightweight assessment?

D.2. Section 26.1.3 presents both the moral case and the business case for data ethics. Some critics argue that framing ethics in business terms is itself a form of ethics-washing — that it reduces moral obligations to economic calculations. Others argue that the business case is the only way to get organizational buy-in. Write a 500-word essay evaluating both positions. Where do you come down, and why?

D.3. Dr. Adeyemi states that "context shapes behavior more powerfully than character" (Section 26.2.1), citing Milgram and Zimbardo. Apply this insight to data ethics programs. If individual virtue is insufficient, what organizational structures are necessary to produce ethical behavior consistently? Conversely, can organizational structures alone produce ethical behavior without individual commitment? Where is the balance?

D.4. The chapter discusses ethics committees as "stakeholder proxies" — bodies that represent the interests of people who are not in the room (Section 26.2.1). Critique this concept. Can an internal committee genuinely represent external stakeholders? What are the limits of proxy representation? Propose at least one alternative or supplementary mechanism for ensuring that affected communities have genuine voice in data ethics decisions.

Part E: Research & Extension ****

These are open-ended projects for students seeking deeper engagement. Each requires independent research beyond the textbook.

E.1. Comparative Ethics Programs. Research the data or AI ethics programs of three organizations from different sectors (e.g., one technology company, one healthcare organization, one financial institution). Write a 1,500-word comparative analysis evaluating each program against the design principles from Section 26.2. Which program is strongest? Where are the common gaps? What would a best-practice synthesis look like?

E.2. The Ethics-Washing Debate. The concept of ethics-washing is contested. Some scholars argue it is pervasive and undermines genuine ethics efforts. Others argue the term is overused and discourages organizations from starting ethics programs at all. Research at least three academic or industry sources on each side of this debate. Write a 1,000-word analysis presenting both perspectives and your own assessment.

E.3. Ethics Program Design. Design a complete data ethics program for a fictional company of your choice (specify the industry, size, and primary data practices). Your program should include: (a) a governance structure, (b) a committee charter, (c) an ethical risk assessment process, (d) a culture change strategy, (e) an incentive structure, and (f) metrics for evaluating the program's effectiveness. Present your design in a 2,000-word proposal.

Solutions

Selected solutions are available in appendices/answers-to-selected.md.