Chapter 38 Exercises: Ethics of Political Analytics

Exercise 38.1 — The Four Domains Audit (Individual, 45 minutes)

Select a real political campaign from a recent election cycle (federal, state, or local). Using publicly available information (news coverage, FEC filings, academic studies), conduct an ethical audit of the campaign's data practices across the four domains:

  1. Privacy: What data sources did the campaign appear to use? Were any sensitive categories (health, religion, financial distress) reportedly involved?
  2. Manipulation: Did the campaign use negative advertising or suppression-adjacent tactics? How would you classify them on the persuasion-manipulation spectrum?
  3. Representation: Did the campaign's polling and targeting appear to include or exclude particular demographic communities? Was there any evidence of differential treatment?
  4. Accountability: Did the campaign or its contractors face any regulatory, professional, or journalistic accountability for their data practices?

Write a 1,000-word memo summarizing your findings and identifying the most significant ethical concerns you observed.

Exercise 38.2 — The Dual-Use Design Challenge (Pairs, 60 minutes)

With a partner, complete the following:

Part A: Design a legitimate voter mobilization model for a hypothetical Democratic candidate in a competitive House district. Specify: - What data inputs you would use - What the model predicts - How the output would be used in field operations - What ethical constraints you would build into the design

Part B: Now, with as few changes as possible, describe how the same model architecture could be repurposed for voter suppression. What inputs change? What outputs change? What stays the same?

Part C: Compare your two models and identify: - The minimum change required to flip from mobilization to suppression - What design choices, if any, would make the mobilization version more resistant to suppression repurposing - Whether you believe such resistance is technically feasible or inherently limited

Discuss your findings as a class. Pay attention to disagreements about where the ethical line falls.

Exercise 38.3 — AAPOR Code Application (Individual, 30 minutes)

Read the following fictional poll release:

"A new poll by Benchmark Analytics shows that 67% of voters support the Protect Our Children Online Act. The poll was conducted among registered voters in the state of Ohio."

Using the AAPOR Disclosure Principles, identify every required element that is missing from this release. Write the minimum additional disclosure that would bring this release into compliance with the Code. Then write a paragraph explaining what a journalist should be skeptical about based on the information that remains missing.

Exercise 38.4 — The Vivian Park Decision Tree (Individual, 45 minutes)

Review Vivian Park's dilemma in section 38.8 and the three options she considered.

  1. Map out a complete decision tree for Vivian's choices, including options she may not have considered (e.g., consulting AAPOR's ethics committee, negotiating contractual language about future poll releases, requiring the leading-question result not be publicly attributed to Meridian without the methodological note).

  2. For each branch of the tree, identify: - The likely outcome for Meridian's client relationship - The outcome for Meridian's professional reputation - The outcome for public information about the ballot measure - The ethical evaluation under the AAPOR Code

  3. Which option do you believe Vivian should have chosen? Defend your answer in 300 words, specifically engaging with the tension between client confidentiality and public information accuracy.

Exercise 38.5 — Dark Patterns Identification Lab (Group, 60 minutes)

Materials needed: Access to political campaign email archives (examples are available at politicalemails.org and similar archiving sites).

In groups of three, review 15-20 political fundraising or GOTV emails from the 2022 or 2024 election cycles. For each email, evaluate:

  1. Does the email use accurate representations of its purpose?
  2. Does the interface design make desired actions (e.g., donating) easier than undesired actions (e.g., unsubscribing)?
  3. Does the email use social proof that is fabricated, exaggerated, or misleading?
  4. Does the email impersonate an authoritative source?
  5. Is the email attempting to create false urgency?

Build a simple scoring rubric and rate each email from 0 (no dark pattern elements) to 5 (multiple dark pattern elements). Share your ratings as a class and discuss: - Which dark patterns were most common? - Were dark patterns more prevalent from one party or one type of campaign? - At what score would you say a communication crosses from aggressive to ethically problematic?

Exercise 38.6 — Professional Ethics Comparison Essay (Individual, 90 minutes)

Write a 1,500-word comparative essay addressing the following prompt:

Political analytics draws practitioners from academic political science, data journalism, and commercial data science. Each tradition has developed its own professional ethics framework. Evaluate the strengths and limitations of each framework for governing political analytics practice, and argue for a hybrid ethical framework that synthesizes the best elements of each. Address specifically: (a) what the hybrid framework would require regarding disclosure of analytical methods; (b) how it would handle the client confidentiality vs. public interest tension; (c) what enforcement mechanisms would be feasible.

Your essay should engage with at least three specific provisions of the AAPOR Code, at least one specific provision of the ACM Code of Ethics, and at least one published case study of an ethical controversy in political analytics.

For each of the following campaign data practices, apply the "consent of the governed" test described in section 38.6.1: Would the targeted voter recognize this tactic as legitimate democratic competition if they knew about it?

  1. Purchasing location data from a mobile data broker to identify voters who attend evangelical churches and targeting them with religious-liberty messaging
  2. Using voter file turnout history to concentrate GOTV calls on low-propensity registered supporters
  3. Running digital ads that accurately quote an opponent's position on gun control, targeted to voters with hunting-related consumer purchases
  4. Targeting opposition-leaning voters with accurate information about polling place changes that makes voting more logistically inconvenient (without misrepresentation, just emphasis)
  5. Using a predictive model of demographic characteristics to infer likely racial identity and routing voters away from a candidate who has made racially coded appeals

For each practice, identify: (a) your answer to the consent test, (b) whether the practice is legal, (c) whether legality and ethics converge or diverge, and (d) what additional information would change your evaluation.