Case Study 12.2: Cambridge Analytica — Data-Driven Propaganda and the 2016 Elections

Overview

In March 2018, investigative reports in The Guardian, The New York Times, and The Observer simultaneously published accounts of Cambridge Analytica (CA), a political data firm that had harvested the personal data of approximately 87 million Facebook users without their informed consent and used this data to construct psychographic profiles for targeted political advertising. The revelations triggered the largest privacy scandal in social media history, congressional hearings on both sides of the Atlantic, a $5 billion FTC fine against Facebook, Cambridge Analytica's own collapse, and a fundamental shift in public understanding of the relationship between social media data, political advertising, and democratic governance.

The Cambridge Analytica case is at the intersection of several themes in this chapter: data-driven propaganda, psychographic targeting, the evolution of propaganda from mass broadcast to individualized digital delivery, and the ethics of political persuasion. It also illustrates the significant gap between what was claimed — by Cambridge Analytica itself and in much subsequent media coverage — and what the evidence actually establishes.

Learning Objectives

By engaging with this case study, students will be able to: 1. Explain the OCEAN psychographic model and how it can be operationalized through behavioral data 2. Distinguish between what Cambridge Analytica claimed to have done and what the available evidence establishes 3. Analyze CA's methods as an evolution of classical propaganda techniques into data-driven form 4. Evaluate the legal and ethical violations involved and assess the adequacy of regulatory responses 5. Apply the propaganda-persuasion distinction to data-driven political advertising

Background: The Academic Origins

The OCEAN Model and Behavioral Data

The Cambridge Analytica story begins not in a political consulting firm but in an academic laboratory at Cambridge University.

The OCEAN Model: Psychologists have used the OCEAN (or "Big Five") model — Openness to experience, Conscientiousness, Extraversion, Agreeableness, and Neuroticism — as a framework for describing personality variation since the 1980s. The model has substantial empirical support: OCEAN dimensions predict a range of important life outcomes including job performance, relationship quality, health behaviors, and political attitudes.

Kosinski et al. (2013): A landmark paper by Michal Kosinski, David Stillwell, and Thore Graepel in Proceedings of the National Academy of Sciences demonstrated that OCEAN personality profiles could be predicted with high accuracy from Facebook "likes" — the pages, posts, and public content that users had voluntarily liked. The researchers found that: - Using 10 Facebook likes, their model outperformed the average work colleague in personality assessment - Using 70 likes, it outperformed friends and family - Using 300 likes, it outperformed spouses

The study used voluntarily shared public data from 58,000 Facebook users who had explicitly consented to participate in a personality study, and was conducted for academic research purposes with full ethics review.

The political connection: The Kosinski et al. finding had obvious implications beyond personality psychology — if personality could be predicted from social media behavior, and if personality predicts political attitudes and voting behavior, then social media data could be used to construct political targeting models of unprecedented precision.

Cambridge Analytica: The Company

Structure and Funding

Cambridge Analytica was established in 2013 as the US political operations arm of SCL Group (Strategic Communication Laboratories), a UK-based defense and intelligence contractor that had previously worked on psychological operations (PSYOP) for military and government clients in multiple countries. SCL had worked on election influence campaigns across Africa and South Asia.

Cambridge Analytica's primary backer was Robert Mercer, a wealthy Republican donor and hedge fund billionaire, with board representation from Steve Bannon, who later served as chief strategist for the Trump White House.

The Data Harvest

The data that Cambridge Analytica used was collected not through academic research but through a separate operation:

Aleksandr Kogan and "This Is Your Digital Life": Aleksandr Kogan, a Cambridge University psychology researcher with connections to both Kosinski's lab and Russian academic institutions, developed a Facebook application called "This Is Your Digital Life" — a personality quiz that paid users for completing a psychographic survey. Crucially, the app exploited Facebook's API (application programming interface) permissions that, at the time, allowed apps to access not just the data of users who explicitly installed the app but also the data of their Facebook friends — without those friends' knowledge or consent.

Kogan's app collected data from approximately 270,000 users who explicitly installed it — and leveraged that access to scrape data from their network of Facebook friends, reaching an estimated 87 million Facebook users in total.

Kogan sold this data to Cambridge Analytica. This transfer violated Facebook's platform policies (which prohibited commercial transfer of user data collected through such apps) and subsequently became the focus of regulatory investigation.

CA's Data Usage

Cambridge Analytica used the harvested data to build psychographic targeting models for political clients. The company claimed to have built individual OCEAN profiles for over 230 million American adults, enabling the delivery of precisely tailored political advertising to individual voters based on their psychological characteristics.

The operational concept: different messaging for different personality profiles. A voter scoring high on Neuroticism and low on Agreeableness would receive different political content than a voter scoring high on Openness and Conscientiousness — content calibrated to resonate with their specific psychological profile rather than designed for a mass audience.

What Cambridge Analytica Actually Did: Evidence vs. Claims

The Claims

Cambridge Analytica was extraordinarily boastful about its capabilities. Alexander Nix, CA's CEO, presented the company's capabilities to political clients and at industry conferences in terms that implied near-miraculous precision in voter manipulation:

• That CA had constructed individual psychological profiles for virtually all American voters
• That these profiles enabled delivery of precisely targeted political content that changed voting behavior
• That CA had been primarily responsible for Donald Trump's 2016 electoral victory and the Brexit Leave campaign's success
• That CA's targeting capabilities were so sophisticated that targeted voters were unaware they were being manipulated

What the Evidence Actually Shows

Independent researchers, academic studies, and investigations by the UK Information Commissioner's Office, the US FTC, and the Special Counsel's office have produced a more ambiguous picture:

The data harvest was real and legally problematic: The scraping of 87 million Facebook profiles without adequate consent is not disputed. Facebook violated its own platform policies by allowing this data to be scraped; Kogan violated those policies by transferring it; Cambridge Analytica violated norms of research ethics and potentially law by using it for commercial political purposes. These facts are established.

CA's proprietary data was likely less powerful than claimed: Academic researchers who examined CA's methodology were skeptical that the firm's proprietary psychographic data provided meaningful targeting advantages over standard demographic targeting. The Kosinski et al. academic finding, while robust, did not straightforwardly translate into the commercial product CA claimed to have built. The firm may have substantially overstated the precision and power of its profiles for commercial purposes.

Causal attribution is very difficult: Establishing that Cambridge Analytica's targeting caused specific electoral outcomes requires controlling for all the other factors that influenced those outcomes — a nearly impossible research design problem in real-world elections. No published peer-reviewed study has established that CA's targeting had a specific, measurable effect on 2016 election outcomes.

The Trump campaign used CA, but also many other vendors: The Trump campaign purchased CA's services, but it also used data and targeting from the Republican National Committee, other data firms, and its own internal analytics. Attributing campaign success specifically to CA's contribution versus other factors is not possible from available evidence.

Brexit: While CA worked for the Leave campaign, subsequent investigations found that CA's role in the Brexit campaign was more limited than claimed. The Information Commissioner's Office investigation found insufficient evidence that CA's data was used in the Brexit referendum campaign.

What was admitted in the Nix recordings: In secretly recorded footage obtained by Channel 4 News (2018), CEO Alexander Nix boasted about using entrapment operations (sending attractive women and wealthy men to meet with politicians), fabricating "deep state" opposition material, coordinated fake grassroots campaigns, and other operations that went far beyond data analytics. These boasts may themselves have been exaggerated for sales purposes, but they described clearly illegal and deeply unethical activities.

Cambridge Analytica as Evolved Propaganda

Applying the IPA Framework

Cambridge Analytica's operations represent an evolution of classical propaganda techniques into data-driven form:

Testimonial, evolved: Rather than using a specific celebrity, CA's targeting model identifies which type of credibility signal (expert authority, peer testimonial, values alignment) is most persuasive for each personality profile and delivers the appropriate credibility framing. Psychographic testimonial targeting.

Bandwagon, evolved: Social proof messaging ("join the millions of [personality-similar people] who...") can be precisely calibrated to each individual's social conformity orientation — strong bandwagon appeals for high-conformity personalities, independence-emphasizing appeals for low-conformity personalities.

Card stacking, evolved: A/B testing at scale enables identification of which selective presentations of accurate information are most persuasive for each audience segment, mechanizing and optimizing the selective presentation technique.

Fear appeals: CA's work reportedly included identification of which fear appeals — fear of crime, economic anxiety, cultural change — were most activating for specific personality profiles, enabling precisely targeted fear-based messaging.

Plain Folks and Transfer, evolved: Visual and tonal content can be A/B tested and matched to personality profiles — more rustic, working-class imagery for plain folks-responsive profiles; more aspirational, achievement-oriented imagery for others.

New Dimensions Not in the IPA Framework

Data-driven propaganda also involves capabilities that exceed the IPA framework's categories:

Precision personalization: The IPA techniques were designed for mass audiences; CA's approach (in principle) targets individuals. The ethical problem of personalization is distinct from mass manipulation: it enables calibrated targeting of specific vulnerabilities unique to specific individuals.

Scale: Simultaneously delivering personally tailored propaganda messages to hundreds of millions of individuals is quantitatively different from broadcast propaganda in ways that may constitute a qualitative change.

Opacity: Political advertising delivered as dark ads to specific individuals is invisible to the general public, to researchers, and to regulators in ways that broadcast propaganda is not. This opacity creates a novel accountability problem.

The feedback loop: Digital advertising generates behavioral data (click-through rates, engagement, sharing) that enables continuous optimization. The propaganda message is not fixed but continuously refined based on real-time behavioral feedback — a level of adaptive optimization unavailable to historical propagandists.

Legal Outcomes and Regulatory Responses

Facebook's $5 Billion FTC Fine (2019)

The Federal Trade Commission imposed a $5 billion fine on Facebook — the largest privacy fine in US history at the time — for violating a 2012 consent decree regarding user data privacy. The fine was criticized by privacy advocates as insufficient relative to Facebook's revenues and insufficient to deter future violations.

UK Information Commissioner's Office

The ICO investigated Cambridge Analytica and issued a £500,000 fine against Facebook (the maximum available under pre-GDPR law). The ICO's investigation of CA's activities was more extensive but was impeded by CA's collapse and the destruction of evidence. The ICO's full investigation report (2020) documented the inadequate regulatory framework for political data use in the UK.

Cambridge Analytica's Collapse

Cambridge Analytica ceased operations in May 2018, following the press revelations. The company's collapse was largely driven by the reputational damage and legal costs of the scandal, though its principals subsequently formed successor companies that continued similar operations.

US Congressional Hearings

Congressional hearings featuring Mark Zuckerberg (April 2018) established the scale of the data harvest and Facebook's negligence in allowing it, but produced relatively limited legislative action in the United States. The absence of comprehensive US federal privacy legislation remained unchanged.

GDPR and European Regulatory Framework

The EU's General Data Protection Regulation (GDPR), which came into full effect in May 2018 — two months after the Cambridge Analytica revelations — would have prevented many of the data practices at the center of the scandal. GDPR requires explicit, informed consent for data processing; prohibits the collection of data from third parties without their consent; and creates significant penalties for violations.

Implications for Democratic Governance

The Dark Ads Problem

Cambridge Analytica's political advertising was delivered as "dark ads" — advertisements visible only to their targeted recipients. This opacity has several democratic implications:

Research gap: Researchers cannot study the content of dark political advertising systematically, making it impossible to assess the accuracy, fairness, or manipulation techniques employed.

Regulatory gap: Advertising that is not publicly visible cannot be reviewed for compliance with political advertising laws.

Accountability gap: Voters cannot be exposed to the range of political messages being sent by campaigns — different voters receive entirely different pictures of the same campaign. This fragments the shared informational basis for democratic deliberation.

The Consent Problem

The Cambridge Analytica case raised fundamental questions about the nature of consent in the digital economy. Facebook users consented to share their data with Facebook through the terms of service — but: - They did not consent to that data being used to construct psychographic profiles - They did not consent to those profiles being used for political targeting - 87 million of them did not consent at all — their data was collected from friends' app usage without any individual consent

The concept of "informed consent" requires that individuals understand what they are consenting to. The Cambridge Analytica case demonstrated that generic terms of service consent does not constitute informed consent for specific data uses — a principle now codified in GDPR.

Discussion Questions

1. If Cambridge Analytica's actual influence on electoral outcomes was more limited than claimed — if psychographic targeting did not demonstrably change votes — does this reduce the ethical seriousness of its methods? Or is the moral wrong in the attempt to manipulate, independent of its effectiveness?

2. The case illustrates a significant gap between what Cambridge Analytica claimed (precision psychological manipulation at scale) and what the evidence established (uncertain effectiveness, significant data violations). How should we evaluate the Cambridge Analytica narrative now that we know the claims were substantially overstated? Does this change the regulatory response it triggered?

3. Cambridge Analytica's methods represent an evolution of classical propaganda techniques (bandwagon, testimonial, fear appeals, card stacking) into data-driven form. Does the fact that the underlying techniques are old change your assessment of the novelty or severity of the threat?

4. The GDPR would have prevented many of Cambridge Analytica's data practices. The United States has no equivalent federal privacy law. What are the strongest arguments for and against comprehensive federal privacy legislation in the United States?

5. Individual political micro-targeting — delivering different political messages to different people based on their psychological profiles — has been criticized as fundamentally incompatible with democratic deliberation, which requires a shared informational basis for collective judgment. Do you agree? Is there a version of political personalization that could be ethically acceptable?

Technical Note for Code Examples

The code/case-study-code.py file for Chapter 12 demonstrates: 1. A simplified simulation of personality-based message targeting (using the OCEAN dimensions) 2. Analysis of how different propaganda techniques map to different personality profiles 3. A mock A/B testing framework showing how political message variants are optimized 4. Visualization of the targeting architecture

These demonstrations are for educational purposes and use synthetic data.

Key Sources

• Cadwalladr, Carole. "The Great British Brexit Robbery." The Observer, May 7, 2017.
• Rosenberg, Matthew, Nicholas Confessore, and Carole Cadwalladr. "How Trump Consultants Exploited the Facebook Data of Millions." New York Times, March 17, 2018.
• Wylie, Christopher. Mindfck: Cambridge Analytica and the Plot to Break America*. Random House, 2019. (Insider account; read critically)
• UK Information Commissioner's Office. Investigation into the use of data analytics in political campaigns. ICO, November 2018.
• Kosinski, M., Stillwell, D., and Graepel, T. "Private traits and attributes are predictable from digital records of human behavior." PNAS 110(15), 2013.
• Nix, Alexander. Promotional presentations at Concordia Summit (2016) and other venues — documented in multiple journalistic accounts.
• Channel 4 News Investigations Team. "Cambridge Analytica Uncovered." March 2018. (Includes secretly recorded footage of Nix's statements)
• Federal Trade Commission. "FTC Imposes $5 Billion Penalty and Sweeping New Privacy Restrictions on Facebook." Press release, July 24, 2019.