Data Versus Democracy Read online

  How Big Data Algorithms

  Shape Opinions and Alter

  the Course of History

  —

  Kris Shaffer

  DATA VERSUS DEMOCRACY

  HOW BIG DATA ALGORITHMS

  SHAPE OPINIONS AND ALTER THE

  COURSE OF HISTORY

  Kris Shaf fer

  Data versus Democracy: How Big Data Algorithms Shape Opinions and

  Alter the Course of History

  Kris Shaffer

  Colorado, USA

  ISBN-13

  (pbk):

  978-1-4842-4539-2 ISBN-13

  (electronic):

  978-1-4842-4540-8

  https://doi.org/10.1007/978-1-4842-4540-8

  Copyright © 2019 by Kris Shaffer

  This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or

  part of the material is concerned, specifically the rights of translation, reprinting, reuse of

  illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way,

  and transmission or information storage and retrieval, electronic adaptation, computer

  software, or by similar or dissimilar methodology now known or hereafter developed.

  Trademarked names, logos, and images may appear in this book. Rather than use a trademark

  symbol with every occurrence of a trademarked name, logo, or image we use the names, logos,

  and images only in an editorial fashion and to the benefit of the trademark owner, with no

  intention of infringement of the trademark.

  The use in this publication of trade names, trademarks, service marks, and similar terms, even if

  they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights.

  While the advice and information in this book are believed to be true and accurate at the date of

  publication, neither the authors nor the editors nor the publisher can accept any legal

  responsibility for any errors or omissions that may be made. The publisher makes no warranty,

  express or implied, with respect to the material contained herein.

  Managing Director, Apress Media LLC: Welmoed Spahr

  Acquisitions Editor: Shiva Ramachandran

  Development Editor: Laura Berendson

  Coordinating Editor: Rita Fernando

  Cover designed by eStudioCalamar

  Distributed to the book trade worldwide by Springer Science+Business Media New York, 233

  Spring Street, 6th Floor, New York, NY 10013. Phone 1-800-SPRINGER, fax (201) 348-4505, e-mail

  [email protected], or visit www.springeronline.com. Apress Media, LLC is a California

  LLC and the sole member (owner) is Springer Science + Business Media Finance Inc (SSBM

  Finance Inc). SSBM Finance Inc is a Delaware corporation.

  For information on translations, please e-mail [email protected], or visit http://www.apress.

  com/rights-permissions.

  Apress titles may be purchased in bulk for academic, corporate, or promotional use. eBook

  versions and licenses are also available for most titles. For more information, reference our Print

  and eBook Bulk Sales web page at http://www.apress.com/bulk-sales.

  Any source code or other supplementary material referenced by the author in this book is

  available to readers on GitHub via the book's product page, located at www.apress.

  com/9781484245392. For more detailed information, please visit http://www.apress.com/

  source-code.

  Printed on acid-free paper

  Blessed are the peacemakers.

  Contents

  About the Author vii

  Acknowledgments ix

  Introduction: From Scarcity to Abundance xi

  Part I:

  The Propaganda Problem 1

  Chapter 1: Pay Attention: How Information Abundance Affects the

  Way We Consume Media 3

  Chapter 2: Cog in the System: How the Limits of Our Brains

  Leave Us Vulnerable to Cognitive Hacking 19

  Chapter 3: Swimming Upstream: How Content Recommendation

  Engines Impact Information and Manipulate

  Our Attention 31

  Part II:

  Case Studies 45

  Chapter 4: Domestic Disturbance: Ferguson, GamerGate, and

  the Rise of the American Alt-Right 47

  Chapter 5: Democracy Hacked, Part 1: Russian Interference and

  the New Cold War 67

  Chapter 6: Democracy Hacked, Part 2: Rumors, Bots, and Genocide

  in the Global South 91

  Chapter 7: Conclusion: Where Do We Go from Here? 109

  Index 117

  About the Author

  Kris Shaffer, PhD, is a data scientist and Senior

  Computational Disinformation Analyst for New

  Knowledge. He coauthored “The Tactics and

  Tropes of the Internet Research Agency,” a report

  prepared for the United States Senate Select

  Committee on Intelligence about Russian inter-

  ference in the 2016 U.S. presidential election on

  social media. Kris has consulted for multiple U.S.

  government agencies, nonprofits, and universities

  on matters related to digital disinformation, data

  ethics, and digital pedagogy.

  In a former (professional) life, Kris was an academic and digital humanist. He

  has taught courses in music theory and cognition, computer science, and digi-

  tal studies at Yale University, the University of Colorado Boulder, the

  University of Mary Washington, and Charleston Southern University. He

  holds a PhD from Yale University.

  Acknowledgments

  How do you write the acknowledgments section for a book like this?

  If researching and writing this book has taught me anything, it’s that some

  things are best said in private and, if possible, in person. So in lieu of public

  acknowledgments, I have decided to give a personal, hand-written note to

  those who educated, inspired, or otherwise helped me on this project. And

  hopefully, that note will be delivered in person and accompanied by a drink or

  a meal.

  However, there is one group of people too large to thank individually, but

  whose influence and motivation have been immeasurable. To all my students

  over the years, who have inspired me with their hard work, their brilliance,

  and their desire to make the world a better place, I thank you. May each of

  you make your own dent in the universe and nudge humanity, even a little bit,

  in the right direction.

  Introduction: From

  Scarcity to

  Abundance

  A Brief History of Information

  and the Propaganda Problem

  As long as we’ve had information, we’ve had disinformation. As long as we’ve

  had advertising, we’ve had attempts at “psychographic profiling.” Ever since

  the invention of the printing press, we’ve had concerns about the corrupting

  influence of mass media. But there are some things that are new in the past

  decade. Information is abundant in a way we couldn’t conceive of just a decade

  or two ago, and the new science of recommendation engines—algorithmically

  selected content, based on personal data profiles—dominates the modern

  media landscape. In this
introduction, we will clear away misconceptions and

  focus on the heart of the problem—the intersection of information abun-

  dance, human psychology, user data profiling, and media recommendation

  algorithms. This is where propaganda finds its way into modern society.

  The Lay of the Land

  Have you ever used a search engine to find a stock photo? Maybe you needed

  a slick header for your blog post, some scenery for your family holiday letter,

  a background for an office party flyer. The results can be pretty good, espe-

  cially if you’re on a professional stock image site (and know how to choose

  your search terms).

  But have you ever used a regular search engine to find images of something

  generic? Try it sometime. Search for images of doctor, then nurse. Professor, then teacher. What do you see?

  Chances are you found some pretty stark stereotypes. White-haired profes-

  sors, wearing tweed, lecturing in front of a chalkboard. Teachers also in front

  of chalkboards, smiling at their eager pupils. Doctors in white coats, deftly

  xi

  Introduction: From Scarcity to Abundance

  wielding their stethoscopes or making notes on their patients’ charts. Nurses

  in blue scrubs, also masters of their charts and scopes. You get the picture.

  Walk through a school, a university, a hospital, a general practitioner’s office,

  though, and you’ll see little that matches these images. Most schools and uni-

  versities abandoned chalk long ago, in favor of dry erase boards and electronic

  projectors. And lecturing to seas of students in rows is increasingly rare,

  particularly with young students. And, by the way, all of these professionals

  tend to dress less formally, and certainly more diversely, than the subjects of

  these search result images.

  Search engines don’t give us reality. They give us the results we expect to see.

  Using a combination of human programming, data from user interaction, and

  an ever-repeating feedback loop of the two, the results of these searches

  gradually become more like the generalizations in our minds. The stereotypes

  we hold in our minds determine what we click on, and those clicks form the

  raw data used by the search engine’s algorithms. These algorithms, in turn,

  form generalizations of expected user behavior, based on our collective clicks,

  and serve up the results we’re most likely to click on, based on that data. We

  perceive, we generalize, we search, we click, the machine perceives (the

  clicks), the machine generalizes, the machine returns results.

  It doesn’t end there. Those search results get used all over the web and in

  print (isn’t that why we were searching in the first place?). Those images

  become part of the backdrop of our view of the world and further fuel the

  generalizations formed by our mind. Thus forms an endless loop: human per-

  ception → human generalization → human behavior → machine perception

  → machine generalization → machine behavior → human perception →

  human behavior … And in each turn through the feedback loop, the stereo-

  type gets more stereotypical. Reality is lost. Though, because the stereotypes

  become part of our media landscape, in a very real sense, reality is also formed.

  But did you notice something else strange about those image results?

  How many of those doctors were men, and how many were women? What

  about the nurses? According to The Wall Street Journal, 32% of doctors in

  the United States in 2012 were women, and the proportion is rising. 1 Is that

  the percentage you saw in your search results? According to the National

  Center for Education Statistics, 49% of tenure-track university faculty and

  57% of non-tenure-track faculty in the United States are women.2 How did

  your professor search results compare?

  1 Josh Mitchell, “Women Notch Progress,” The Wall Street Journal, published December 4,

  2012, www.wsj.com/articles/SB10001424127887323717004578159433220839020.

  2 “Quick Take: Women in Academia,” Catalyst, published October 20, 2017, www.catalyst.

  org/knowledge/women-academia.

  xi i

  Introduction: From Scarcity to Abundance

  Chances are your search results were more stereotype than reality. That’s

  partly our brains’ fault. Our brains make generalizations about what we per-

  ceive in the world, and those generalizations allow us to make predictions

  about the world that help us interact with it more efficiently. Cognitive scien-

  tists have also found that when we form generalizations—called schemas—we

  tend to define those schemas, in part, by contrast with other schemas. In

  other words, we emphasize their differences, often making them more distinct

  from each other in our minds than they are in reality. While this method

  of defining ideas and categories in our mind is usually helpful, it sometimes

  works against us by reinforcing the bias of our environment, including the

  (stereotype-ridden) media we encounter. And when the feedback loop of

  human generalizations, machine generalizations, and media representation

  goes online, that bias gets propagated and reinscribed at near light speed.

  It’s all connected—our media, our memory, our identity, our society. The way

  we interact with the world is directly influenced by the “mental map” we have

  of the world—what’s real, what’s not, and where it all belongs. And while that

  mental map is primarily formed in light of our experiences (with a little help

  from hundreds of thousands of years of human evolution), that map is increas-

  ingly influenced by the media that we consume.

  I say “increasingly” not because something in our brains has changed, but

  because our media landscape has changed so drastically in the last century—

  even the last decade. We have moved from information scarcity to informa-

  tion abundance, from media desert to media ubiquity. And for most—though,

  importantly, not all—humans on this planet, our access to the information

  and media that exists has expanded just as rapidly. Our lived experiences are

  increasingly mediated through media.

  The Limits of Attention

  But one important thing has not changed: the limits of the human body,

  including the brain. Sure, infant mortality3 and life expectancy4 have improved in most societies over the past century, and quality of life has improved for

  many as the result of scientific and humanistic advancement. But the human

  cognitive system—the interaction of the brain and the body, memory and the

  3 “Infant Mortality,” World Health Organization Global Health Observatory (GHO) Data,

  accessed February 5, 2019, www.who.int/gho/child_health/mortality/neonatal_

  infant_text/en/.

  4 “Life Expectancy,” World Health Organization Global Health Observatory (GHO) Data,

  accessed February 5, 2019, www.who.int/gho/mortality_burden_disease/life_

  tables/situation_trends_text/en/.

  xiv

  Introduction: From Scarcity to Abundance

  senses—took its more-or-less modern form tens of thousands of years ago. 5

  The amount of information our brains can hold has not changed, nor has the

  limits of conscious attention.

  And with all the media clamoring for our attention, that attent
ion has become

  our most precious—if also our most overlooked—resource.

  Let’s step back to the world of search engines for a moment. They can be

  hacked. I don’t mean a full-on security breach (though that’s certainly possi-

  ble). I mean they can be manipulated by highly motivated users. Just as what

  people click on determines (in part) the search result rankings, the terms that

  people search for determine (in part) the terms that pop up on autocomplete

  as you type. (For some fun, go to Google and search for your home country/

  state: “Why is [Colorado] so” and see how Google thinks you might want to

  complete that search.) If the searches typed into a search engine can deter-

  mine the autocomplete terms, then a group of people willing to put the time

  in can search for the same thing over and over again until it dominates the

  autocomplete results, increasing the number of people who see it, are influ-

  enced by it, and click on it.

  This very thing happened in 2016. Members of the so-called alt-right move-

  ment (right-wing extremists, often affiliated with groups espousing hateful

  views like white supremacy, anti-Semitism, and hyper-masculine antifeminism)

  successfully manipulated Google’s autocomplete to suggest racist searches

  and pro-alt-right messages.6 When this manipulation was brought to Google’s

  attention, they responded with changes to the system. But no system is com-

  pletely impervious to hacking.

  Not even the human mind.

  Cognitive Hacking

  Any system that draws conclusions based on statistics, in part or in whole,

  can be gamed by a manipulation of the statistics. That’s how the human brain

  works—over time, the things we perceive are aggregated together into gen-

  eralized schemas, which are constantly changing with new information. (The

  more established schemas change more slowly, of course.) By altering the

  statistical input of the brain, a “hacker” can impact the schemas our brain

  5 Though scientists still debate just how many tens of thousands. See Erin Wayman, “When

  Did the Human Mind Evolve to What It Is Today?,” Smithsonian Magazine, published June

  25, 2012, www.smithsonianmag.com/science-nature/when-did-the-human-mind-

  evolve-to-what-it-is-today-140507905/.

  6 Olivia Solon and Sam Levin, “How Google’s Search Algorithm Spreads False Information