Following the 2016 presidential election in the United States, a major scandal broke at Facebook that revealed a new frontier of machine learning in American electoral politics. Multiple investigative reports indicated that Cambridge Analytica (CA), a UK-based data analytics firm backed by conservative funding from the United States, had acquired access to the personal data of more than 50 million Facebook users without authorization. This data was used to “identify the personalities of American voters and influence their behavior” through targeted “psychographic” digital ads. The ensuing scandal led Cambridge Analytica to close in 2018.
Electoral politics is a high-stakes information arms race to understand voter sentiment and influence voter behavior. Cambridge Analytica correctly recognized that voters’ ongoing use of social media provided thousands of relevant data points that could be mined for insights. In October 2016, Cambridge Analytica’s CEO, Alexander Nix, discussed his firm’s strategy with Sky News, stating that “Cambridge Analytica is a political tech company that is delivering hypertargeted – and hyperpersuasive – messages to people on social media.” CA’s machine learning effort typically began with a detailed survey, often presented as a psychological test. Survey responses were combined with other data – for example, Facebook likes. CA then applied machine learning to that data to “identify clusters of people who care about a particular issue… and [then] nuance the messaging of [an] advert according to how people see the world, according to their personalities.”
Investigations revealing improper data access led CA to close in 2018 despite the firm’s denial of any wrongdoing. Yet the lesson from CA’s strategy is clear: in a social media landscape with ever-growing stores of personal data, short-term backlash against CA’s unethical methods will not prevent other actors from trying out this successful strategy in the future. Competitive advantage in the political big data industry comes from having the most innovative machine learning tools for identifying and influencing voters. In the next 2-10 years, we should expect other players in the politics industry to develop ever more sophisticated methods for acquiring and mobilizing data on voter sentiment.
Despite the risks, it is a smart business strategy for firms like CA to target voters through data mining. To avoid the backlash that led to CA’s closure, other actors in this space must be careful to identify themselves as politically-interested parties and adhere to privacy guidelines. But political actors will find many opportunities within these constraints. Social media users enjoy taking surveys and many have established party preferences. These traits can be used to drive voluntary engagement that will provide data for machine learning algorithms. Algorithms can be designed to provide insights into voter trends by issues, geographies, and demographic segments. In an ideal world, this would enable political platforms to be more responsive to voter needs.
The story of Cambridge Analytica’s success and subsequent closure raises important questions for the social media and politics industries. How should social media platforms be required to report on data collection and advertising related to political campaigns? What mechanisms should be established prevent foreign parties from influencing elections through social media?
The machine learning arms race to drive voter engagement is now a permanent feature of electoral politics around the world. As analytics firms and political organizations innovate new ways to understand and engage with constituents, individual voters must develop a stronger awareness of who is trying to influence them online and how their personal data determines which political messages they see. In the high-stakes game of electoral politics, companies seeking to mine voter data to inform political strategies are incentivized to test the boundaries of ethical data collection. It would be valuable to increase regulatory pressure on social media platforms to build transparency and accountability into their sales of political ad space, so that consumers understand how their personal data may be used to influence their behavior at the polling station.
Murad, Ahmed, Hannah Kuchler, and Matthew Garrahan. “How User Digital Footprints Marked Path to Social Media’s Weaponisation.” Financial Times, March 19, 2018, pp. 3. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/2026778369?accountid=11311, accessed November 12, 2018.
Kevin Granville, “Facebook and Cambridge Analytics: What you need to Know as Fallout Widens,” New York Times, March 19, 2018, https://www.nytimes.com/2018/03/19/technology/facebook-cambridge-analytica-explained.html, accessed November 12, 2018.
“What is Psychographics? Understanding The ‘Dark Arts’ Of Marketing That Brought Down Cambridge Analytica,” CBINSIGHTS, June 7, 2018, https://app-cbinsights-com.prd1.ezproxy-prod.hbs.edu/research/what-is-psychographics/, accessed November 12, 2018.
Tom Cheshire, “Behind the scenes at Donald Trump’s UK digital UK war room,” Sky News, October 22, 2016, https://news.sky.com/story/behind-the-scenes-at-donald-trumps-uk-digital-war-room-10626155, accessed November 12, 2018.
“Cambridge Analytica Closing Operations Following Facebook Data Controversy — 4th Update.” Dow Jones Institutional News, May 2, 2018. ProQuest, http://search.proquest.com.ezp-prod1.hul.harvard.edu/docview/2033607705?accountid=11311, accessed November 12, 2018.