Summary: Chapter 10, Conclusion, Afterword

Chapter 10: The Targeted Citizen: Civic Life

The final chapter discusses the role of Big Data models in US political life. O’Neil focuses on a phenomenon known as microtargeting, in which candidates or interest groups try to focus on the voters who matter most and the issues that matter most to those voters. O’Neil notes the successful deployment of online microtargeting in the 2012 election, when Barack Obama’s campaign team hired data scientists to identify the most “persuadable” voters and prospective donors. Hillary Clinton, O’Neil suggests, took a similar route in her 2016 presidential campaign. When voters learn that they are being given disparate messages, as happened when Mitt Romney’s “47%” speech was leaked, they tend to be resentful and distrustful. Yet most voters never realize the extent to which microtargeting is practiced nor the extent to which it gives some swing-state voters a greater say in the nation’s politics.

Conclusion

In the conclusion, O’Neil sizes up the problems with an increasingly algorithm-driven society by use of an analogy. She points out that the Industrial Revolution created tremendous progress and improved the quality of life for many but at a huge cost to some groups. Poor and working-class children, for instance, became laborers in dangerous and exhausting jobs such as coal mining. Similarly, the rise of large-scale mathematical modeling has brought real benefits but harmed and excluded many in the process. To remedy this, O’Neil calls for greater humility and care on the part of her fellow data scientists, tighter regulations on the collection and use of personal data, and private audits of the algorithms most likely to cause harm. She closes with the thought that predictive models can do substantial good in the world—but only if they are used fairly and transparently, to identify and help people rather than target and exploit them.

Afterword

New in the 2017 edition, the afterword takes stock of the role that data and modeling played in the 2016 election. O’Neil points out that the polls were inaccurate for several reasons, including Americans’ unwillingness to disclose their choice of candidate. She also observes that polls can influence an election by making some candidates seem like they have no chance to win (and thus are not worth voting for) and others appear certain to win (leading to that candidate’s voters staying home). Another feature of the contemporary political climate, O’Neil says, is the use of opaque and poorly supervised algorithms to curate news on social media, fueling the echo-chamber effect many have noted and criticized.

O’Neil also uses the afterword to reiterate her call for greater scrutiny of the Big Data algorithms that increasingly govern commerce, education, and politics. She points out that the creators of such algorithms often lack an incentive to share the details with the public, so the work falls to investigators who must reverse-engineer the algorithm by examining its results. O’Neil opines that experts from many different fields—not just software engineers and data scientists—will have a role to play in building systems that hold the algorithms accountable.

Analysis: Chapter 10, Conclusion, Afterword

When O’Neil examines the political impact of Big Data models and algorithms, she finds much the same landscape of flaws and pitfalls as in the other areas she has examined. Like predatory advertising, political microtargeting focuses only on the people whom it thinks will yield a profit: in the case of election campaigns, this means both a relevant vote and a potential donation. Like recidivism models, it assesses voters based on factors not under their immediate control, such as the state they live in. For this reason, political microtargeting provides both a summary and an illustration of the arguments that O’Neil has been building throughout the book.

The idea that some votes count more than others is a familiar one in US politics, owing to the structure of the Electoral College. What O’Neil points out here is that political microtargeting exacerbates this disparity by showing politicians, in detail, whom they must focus on and whom they can safely ignore. This promotes some states—and some populations within those states—to a position of importance while banishing the rest to irrelevance. The strategy dictated by microtargeting does not always work: in the afterword, O’Neil notes that Hillary Clinton lost two midwestern states by narrow margins after minimal campaigning in those states. Still, the widespread use of microtargeting alarms O’Neil because of its ability to turn ordinary voters into political haves and have-nots.