public/how-to-destroy-surveillance-capitalism.fr.html

   1 <html><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8"><title>Comment détruire le capitalisme de la surveillance</title><meta name="generator" content="DocBook XSL Stylesheets V1.79.1"></head><body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"><div lang="fr" class="article"><div class="titlepage"><div><div><h2 class="title"><a name="index"></a>Comment détruire le capitalisme de la surveillance</h2></div><div><div class="authorgroup"><div class="author"><h3 class="author"><span class="firstname">Cory</span> <span class="surname">Doctorow</span></h3></div></div></div><div><p class="copyright">Copyright © 2020 Cory Doctorow</p></div><div><p class="copyright">Copyright © 2020 Petter Reinholdtsen</p></div><div><div class="legalnotice"><a name="idm18"></a><p>
   2       Comment détruire le capitalisme de la surveillance par Cory Doctorow.
   3     </p><p>
   4       Publié par Petter Reinholdtsen.
   5     </p><p>
   6       <a class="ulink" href="http://www.hungry.com/~pere/publisher/" target="_top">http://www.hungry.com/~pere/publisher/</a>.
   7     </p><p>
   8       ISBN 978-82-93828-XX-X (livre de poche)
   9     </p><p>
  10       ISBN 978-82-93828-XX-X (ePub)
  11     </p><p>
  12       Ce livre est disponible à la vente sur <a class="ulink" href="https://www.lulu.com/" target="_top">https://www.lulu.com/</a>.
  13     </p><p>
  14       <span class="inlinemediaobject"><img src="images/cc-some-rights-reserved.png" align="middle" height="38" alt="Creative Commons, certains droits réservés"></span>
  15     </p><p>
  16       Ce livre est sous licence Creative Commons. Cette licence permet toute
  17 utilisation de ce travail, tant que l'attribution est donnée et qu'aucun
  18 matériel dérivé n'est distribué. Pour plus d'informations sur la licence,
  19 visitez <a class="ulink" href="https://creativecommons.org/licenses/by-nd/4.0/" target="_top">https://creativecommons.org/licenses/by-nd/4.0/</a>.
  20     </p></div></div></div><hr></div><div class="toc"><p><b>Table des matières</b></p><dl class="toc"><dt><span class="sect1"><a href="#the-net-of-a-thousand-lies">The net of a thousand lies</a></span></dt><dt><span class="sect1"><a href="#digital-rights-activism-a-quarter-century-on">Digital rights activism, a quarter-century on</a></span></dt><dt><span class="sect1"><a href="#tech-exceptionalism-then-and-now">Tech exceptionalism, then and now</a></span></dt><dt><span class="sect1"><a href="#dont-believe-the-hype">Don’t believe the hype</a></span></dt><dt><span class="sect1"><a href="#what-is-persuasion">What is persuasion?</a></span></dt><dd><dl><dt><span class="sect2"><a href="#segmenting">1. Segmenting</a></span></dt><dt><span class="sect2"><a href="#deception">2. Deception</a></span></dt><dt><span class="sect2"><a href="#domination">3. Domination</a></span></dt><dt><span class="sect2"><a href="#bypassing-our-rational-faculties">4. Bypassing our rational faculties</a></span></dt></dl></dd><dt><span class="sect1"><a href="#if-data-is-the-new-oil-then-surveillance-capitalisms-engine-has-a-leak">If data is the new oil, then surveillance capitalism’s engine has a leak</a></span></dt><dt><span class="sect1"><a href="#what-is-facebook">What is Facebook?</a></span></dt><dt><span class="sect1"><a href="#monopoly-and-the-right-to-the-future-tense">Monopoly and the right to the future tense</a></span></dt><dt><span class="sect1"><a href="#search-order-and-the-right-to-the-future-tense">Search order and the right to the future tense</a></span></dt><dt><span class="sect1"><a href="#monopolists-can-afford-sleeping-pills-for-watchdogs">Monopolists can afford sleeping pills for watchdogs</a></span></dt><dt><span class="sect1"><a href="#privacy-and-monopoly">Privacy and monopoly</a></span></dt><dt><span class="sect1"><a href="#ronald-reagan-pioneer-of-tech-monopolism">Ronald Reagan, pioneer of tech monopolism</a></span></dt><dt><span class="sect1"><a href="#steering-with-the-windshield-wipers">Steering with the windshield wipers</a></span></dt><dt><span class="sect1"><a href="#surveillance-still-matters">Surveillance still matters</a></span></dt><dt><span class="sect1"><a href="#dignity-and-sanctuary">Dignity and sanctuary</a></span></dt><dt><span class="sect1"><a href="#afflicting-the-afflicted">Afflicting the afflicted</a></span></dt><dt><span class="sect1"><a href="#any-data-you-collect-and-retain-will-eventually-leak">Any data you collect and retain will eventually leak</a></span></dt><dt><span class="sect1"><a href="#critical-tech-exceptionalism-is-still-tech-exceptionalism">Critical tech exceptionalism is still tech exceptionalism</a></span></dt><dt><span class="sect1"><a href="#how-monopolies-not-mind-control-drive-surveillance-capitalism-the-snapchat-story">How monopolies, not mind control, drive surveillance capitalism: The
  21 Snapchat story</a></span></dt><dt><span class="sect1"><a href="#a-monopoly-over-your-friends">A monopoly over your friends</a></span></dt><dt><span class="sect1"><a href="#fake-news-is-an-epistemological-crisis">Fake news is an epistemological crisis</a></span></dt><dt><span class="sect1"><a href="#tech-is-different">Tech is different</a></span></dt><dt><span class="sect1"><a href="#ownership-of-facts">Ownership of facts</a></span></dt><dt><span class="sect1"><a href="#persuasion-works-slowly">Persuasion works… slowly</a></span></dt><dt><span class="sect1"><a href="#paying-wont-help">Paying won’t help</a></span></dt><dt><span class="sect1"><a href="#an-ecology-moment-for-trustbusting">An <span class="quote">« <span class="quote">ecology</span> »</span> moment for trustbusting</a></span></dt><dt><span class="sect1"><a href="#make-big-tech-small-again">Make Big Tech small again</a></span></dt><dt><span class="sect1"><a href="#goto-10">20 GOTO 10</a></span></dt><dt><span class="sect1"><a href="#up-and-through">Up and through</a></span></dt></dl></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="the-net-of-a-thousand-lies"></a>The net of a thousand lies</h2></div></div></div><p>
  22     The most surprising thing about the rebirth of flat Earthers in the 21st
  23 century is just how widespread the evidence against them is.  You can
  24 understand how, centuries ago, people who’d never gained a high-enough
  25 vantage point from which to see the Earth’s curvature might come to the
  26 commonsense belief that the flat-seeming Earth was, indeed, flat.
  27   </p><p>
  28     But today, when elementary schools routinely dangle GoPro cameras from
  29 balloons and loft them high enough to photograph the Earth’s curve — to say
  30 nothing of the unexceptional sight of the curved Earth from an airplane
  31 window — it takes a heroic effort to maintain the belief that the world is
  32 flat.
  33   </p><p>
  34     Likewise for white nationalism and eugenics: In an age where you can become
  35 a computational genomics datapoint by swabbing your cheek and mailing it to
  36 a gene-sequencing company along with a modest sum of money, <span class="quote">« <span class="quote">race
  37 science</span> »</span> has never been easier to refute.
  38   </p><p>
  39     We are living through a golden age of both readily available facts and
  40 denial of those facts. Terrible ideas that have lingered on the fringes for
  41 decades or even centuries have gone mainstream seemingly overnight.
  42   </p><p>
  43     When an obscure idea gains currency, there are only two things that can
  44 explain its ascendance: Either the person expressing that idea has gotten a
  45 lot better at stating their case, or the proposition has become harder to
  46 deny in the face of mounting evidence. In other words, if we want people to
  47 take climate change seriously, we can get a bunch of Greta Thunbergs to make
  48 eloquent, passionate arguments from podiums, winning our hearts and minds,
  49 or we can wait for flood, fire, broiling sun, and pandemics to make the case
  50 for us. In practice, we’ll probably have to do some of both: The more we’re
  51 boiling and burning and drowning and wasting away, the easier it will be for
  52 the Greta Thunbergs of the world to convince us.
  53   </p><p>
  54     The arguments for ridiculous beliefs in odious conspiracies like
  55 anti-vaccination, climate denial, a flat Earth, and eugenics are no better
  56 than they were a generation ago. Indeed, they’re worse because they are
  57 being pitched to people who have at least a background awareness of the
  58 refuting facts.
  59   </p><p>
  60     Anti-vax has been around since the first vaccines, but the early
  61 anti-vaxxers were pitching people who were less equipped to understand even
  62 the most basic ideas from microbiology, and moreover, those people had not
  63 witnessed the extermination of mass-murdering diseases like polio, smallpox,
  64 and measles. Today’s anti-vaxxers are no more eloquent than their forebears,
  65 and they have a much harder job.
  66   </p><p>
  67     So can these far-fetched conspiracy theorists really be succeeding on the
  68 basis of superior arguments?
  69   </p><p>
  70     Some people think so. Today, there is a widespread belief that machine
  71 learning and commercial surveillance can turn even the most fumble-tongued
  72 conspiracy theorist into a svengali who can warp your perceptions and win
  73 your belief by locating vulnerable people and then pitching them with
  74 A.I.-refined arguments that bypass their rational faculties and turn
  75 everyday people into flat Earthers, anti-vaxxers, or even Nazis. When the
  76 RAND Corporation <a class="ulink" href="https://www.rand.org/content/dam/rand/pubs/research_reports/RR400/RR453/RAND_RR453.pdf" target="_top">blames
  77 Facebook for <span class="quote">« <span class="quote">radicalization</span> »</span></a> and when Facebook’s role
  78 in spreading coronavirus misinformation is <a class="ulink" href="https://secure.avaaz.org/campaign/en/facebook_threat_health/" target="_top">blamed on
  79 its algorithm</a>, the implicit message is that machine learning and
  80 surveillance are causing the changes in our consensus about what’s true.
  81   </p><p>
  82     After all, in a world where sprawling and incoherent conspiracy theories
  83 like Pizzagate and its successor, QAnon, have widespread followings,
  84 <span class="emphasis"><em>something</em></span> must be afoot.
  85   </p><p>
  86     But what if there’s another explanation? What if it’s the material
  87 circumstances, and not the arguments, that are making the difference for
  88 these conspiracy pitchmen? What if the trauma of living through
  89 <span class="emphasis"><em>real conspiracies</em></span> all around us — conspiracies among
  90 wealthy people, their lobbyists, and lawmakers to bury inconvenient facts
  91 and evidence of wrongdoing (these conspiracies are commonly known as
  92 <span class="quote">« <span class="quote">corruption</span> »</span>) — is making people vulnerable to conspiracy
  93 theories?
  94   </p><p>
  95     If it’s trauma and not contagion — material conditions and not ideology —
  96 that is making the difference today and enabling a rise of repulsive
  97 misinformation in the face of easily observed facts, that doesn’t mean our
  98 computer networks are blameless. They’re still doing the heavy work of
  99 locating vulnerable people and guiding them through a series of
 100 ever-more-extreme ideas and communities.
 101   </p><p>
 102     Belief in conspiracy is a raging fire that has done real damage and poses
 103 real danger to our planet and species, from epidemics <a class="ulink" href="https://www.cdc.gov/measles/cases-outbreaks.html" target="_top">kicked off by vaccine
 104 denial</a> to genocides <a class="ulink" href="https://www.nytimes.com/2018/10/15/technology/myanmar-facebook-genocide.html" target="_top">kicked
 105 off by racist conspiracies</a> to planetary meltdown caused by
 106 denial-inspired climate inaction. Our world is on fire, and so we have to
 107 put the fires out — to figure out how to help people see the truth of the
 108 world through the conspiracies they’ve been confused by.
 109   </p><p>
 110     But firefighting is reactive. We need fire
 111 <span class="emphasis"><em>prevention</em></span>. We need to strike at the traumatic material
 112 conditions that make people vulnerable to the contagion of conspiracy. Here,
 113 too, tech has a role to play.
 114   </p><p>
 115     There’s no shortage of proposals to address this. From the EU’s <a class="ulink" href="https://edri.org/tag/terreg/" target="_top">Terrorist Content Regulation</a>,
 116 which requires platforms to police and remove <span class="quote">« <span class="quote">extremist</span> »</span>
 117 content, to the U.S. proposals to <a class="ulink" href="https://www.eff.org/deeplinks/2020/03/earn-it-act-violates-constitution" target="_top">force
 118 tech companies to spy on their users</a> and hold them liable <a class="ulink" href="https://www.natlawreview.com/article/repeal-cda-section-230" target="_top">for their
 119 users’ bad speech</a>, there’s a lot of energy to force tech companies
 120 to solve the problems they created.
 121   </p><p>
 122     There’s a critical piece missing from the debate, though. All these
 123 solutions assume that tech companies are a fixture, that their dominance
 124 over the internet is a permanent fact. Proposals to replace Big Tech with a
 125 more diffused, pluralistic internet are nowhere to be found. Worse: The
 126 <span class="quote">« <span class="quote">solutions</span> »</span> on the table today <span class="emphasis"><em>require</em></span> Big
 127 Tech to stay big because only the very largest companies can afford to
 128 implement the systems these laws demand.
 129   </p><p>
 130     Figuring out what we want our tech to look like is crucial if we’re going to
 131 get out of this mess. Today, we’re at a crossroads where we’re trying to
 132 figure out if we want to fix the Big Tech companies that dominate our
 133 internet or if we want to fix the internet itself by unshackling it from Big
 134 Tech’s stranglehold. We can’t do both, so we have to choose.
 135   </p><p>
 136     I want us to choose wisely. Taming Big Tech is integral to fixing the
 137 Internet, and for that, we need digital rights activism.
 138   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="digital-rights-activism-a-quarter-century-on"></a>Digital rights activism, a quarter-century on</h2></div></div></div><p>
 139     Digital rights activism is more than 30 years old now. The Electronic
 140 Frontier Foundation turned 30 this year; the Free Software Foundation
 141 launched in 1985. For most of the history of the movement, the most
 142 prominent criticism leveled against it was that it was irrelevant: The real
 143 activist causes were real-world causes (think of the skepticism when <a class="ulink" href="https://www.loc.gov/law/foreign-news/article/finland-legal-right-to-broadband-for-all-citizens/#:~:text=Global%20Legal%20Monitor,-Home%20%7C%20Search%20%7C%20Browse&amp;text=(July%206%2C%202010)%20On,connection%20100%20MBPS%20by%202015." target="_top">Finland
 144 declared broadband a human right in 2010</a>), and real-world activism
 145 was shoe-leather activism (think of Malcolm Gladwell’s <a class="ulink" href="https://www.newyorker.com/magazine/2010/10/04/small-change-malcolm-gladwell" target="_top">contempt
 146 for <span class="quote">« <span class="quote">clicktivism</span> »</span></a>). But as tech has grown more central
 147 to our daily lives, these accusations of irrelevance have given way first to
 148 accusations of insincerity (<span class="quote">« <span class="quote">You only care about tech because you’re
 149 <a class="ulink" href="https://www.ipwatchdog.com/2018/06/04/report-engine-eff-shills-google-patent-reform/id=98007/" target="_top">shilling
 150 for tech companies</a></span> »</span>) to accusations of negligence (<span class="quote">« <span class="quote">Why
 151 didn’t you foresee that tech could be such a destructive force?</span> »</span>).
 152 But digital rights activism is right where it’s always been: looking out for
 153 the humans in a world where tech is inexorably taking over.
 154   </p><p>
 155     The latest version of this critique comes in the form of <span class="quote">« <span class="quote">surveillance
 156 capitalism,</span> »</span> a term coined by business professor Shoshana Zuboff in
 157 her long and influential 2019 book, <span class="emphasis"><em>The Age of Surveillance
 158 Capitalism: The Fight for a Human Future at the New Frontier of
 159 Power</em></span>. Zuboff argues that <span class="quote">« <span class="quote">surveillance capitalism</span> »</span>
 160 is a unique creature of the tech industry and that it is unlike any other
 161 abusive commercial practice in history, one that is <span class="quote">« <span class="quote">constituted by
 162 unexpected and often illegible mechanisms of extraction, commodification,
 163 and control that effectively exile persons from their own behavior while
 164 producing new markets of behavioral prediction and
 165 modification. Surveillance capitalism challenges democratic norms and
 166 departs in key ways from the centuries-long evolution of market
 167 capitalism.</span> »</span> It is a new and deadly form of capitalism, a
 168 <span class="quote">« <span class="quote">rogue capitalism,</span> »</span> and our lack of understanding of its unique
 169 capabilities and dangers represents an existential, species-wide
 170 threat. She’s right that capitalism today threatens our species, and she’s
 171 right that tech poses unique challenges to our species and civilization, but
 172 she’s really wrong about how tech is different and why it threatens our
 173 species.
 174   </p><p>
 175     What’s more, I think that her incorrect diagnosis will lead us down a path
 176 that ends up making Big Tech stronger, not weaker. We need to take down Big
 177 Tech, and to do that, we need to start by correctly identifying the problem.
 178   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="tech-exceptionalism-then-and-now"></a>Tech exceptionalism, then and now</h2></div></div></div><p>
 179     Early critics of the digital rights movement — perhaps best represented by
 180 campaigning organizations like the Electronic Frontier Foundation, the Free
 181 Software Foundation, Public Knowledge, and others that focused on preserving
 182 and enhancing basic human rights in the digital realm — damned activists for
 183 practicing <span class="quote">« <span class="quote">tech exceptionalism.</span> »</span> Around the turn of the
 184 millennium, serious people ridiculed any claim that tech policy mattered in
 185 the <span class="quote">« <span class="quote">real world.</span> »</span> Claims that tech rules had implications for
 186 speech, association, privacy, search and seizure, and fundamental rights and
 187 equities were treated as ridiculous, an elevation of the concerns of sad
 188 nerds arguing about <span class="emphasis"><em>Star Trek</em></span> on bulletin board systems
 189 above the struggles of the Freedom Riders, Nelson Mandela, or the Warsaw
 190 ghetto uprising.
 191   </p><p>
 192     In the decades since, accusations of <span class="quote">« <span class="quote">tech exceptionalism</span> »</span> have
 193 only sharpened as tech’s role in everyday life has expanded: Now that tech
 194 has infiltrated every corner of our life and our online lives have been
 195 monopolized by a handful of giants, defenders of digital freedoms are
 196 accused of carrying water for Big Tech, providing cover for its
 197 self-interested negligence (or worse, nefarious plots).
 198   </p><p>
 199     From my perspective, the digital rights movement has remained stationary
 200 while the rest of the world has moved. From the earliest days, the
 201 movement’s concern was users and the toolsmiths who provided the code they
 202 needed to realize their fundamental rights.  Digital rights activists only
 203 cared about companies to the extent that companies were acting to uphold
 204 users’ rights (or, just as often, when companies were acting so foolishly
 205 that they threatened to bring down new rules that would also make it harder
 206 for good actors to help users).
 207   </p><p>
 208     The <span class="quote">« <span class="quote">surveillance capitalism</span> »</span> critique recasts the digital
 209 rights movement in a new light again: not as alarmists who overestimate the
 210 importance of their shiny toys nor as shills for big tech but as serene
 211 deck-chair rearrangers whose long-standing activism is a liability because
 212 it makes them incapable of perceiving novel threats as they continue to
 213 fight the last century’s tech battles.
 214   </p><p>
 215     But tech exceptionalism is a sin no matter who practices it.
 216   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="dont-believe-the-hype"></a>Don’t believe the hype</h2></div></div></div><p>
 217     You’ve probably heard that <span class="quote">« <span class="quote">if you’re not paying for the product,
 218 you’re the product.</span> »</span> As we’ll see below, that’s true, if incomplete.
 219 But what is <span class="emphasis"><em>absolutely</em></span> true is that ad-driven Big
 220 Tech’s customers are advertisers, and what companies like Google and
 221 Facebook sell is their ability to convince <span class="emphasis"><em>you</em></span> to buy
 222 stuff. Big Tech’s product is persuasion. The services — social media, search
 223 engines, maps, messaging, and more — are delivery systems for persuasion.
 224   </p><p>
 225     The fear of surveillance capitalism starts from the (correct)  presumption
 226 that everything Big Tech says about itself is probably a lie. But the
 227 surveillance capitalism critique makes an exception for the claims Big Tech
 228 makes in its sales literature — the breathless hype in the pitches to
 229 potential advertisers online and in ad-tech seminars about the efficacy of
 230 its products: It assumes that Big Tech is as good at influencing us as they
 231 claim they are when they’re selling influencing products to credulous
 232 customers. That’s a mistake because sales literature is not a reliable
 233 indicator of a product’s efficacy.
 234   </p><p>
 235     Surveillance capitalism assumes that because advertisers buy a lot of what
 236 Big Tech is selling, Big Tech must be selling something real. But Big Tech’s
 237 massive sales could just as easily be the result of a popular delusion or
 238 something even more pernicious: monopolistic control over our communications
 239 and commerce.
 240   </p><p>
 241     Being watched changes your behavior, and not for the better. It creates
 242 risks for our social progress. Zuboff’s book features beautifully wrought
 243 explanations of these phenomena. But Zuboff also claims that surveillance
 244 literally robs us of our free will — that when our personal data is mixed
 245 with machine learning, it creates a system of persuasion so devastating that
 246 we are helpless before it.  That is, Facebook uses an algorithm to analyze
 247 the data it nonconsensually extracts from your daily life and uses it to
 248 customize your feed in ways that get you to buy stuff. It is a mind-control
 249 ray out of a 1950s comic book, wielded by mad scientists whose
 250 supercomputers guarantee them perpetual and total world domination.
 251   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="what-is-persuasion"></a>What is persuasion?</h2></div></div></div><p>
 252     To understand why you shouldn’t worry about mind-control rays — but why you
 253 <span class="emphasis"><em>should</em></span> worry about surveillance
 254 <span class="emphasis"><em>and</em></span> Big Tech — we must start by unpacking what we mean
 255 by <span class="quote">« <span class="quote">persuasion.</span> »</span>
 256   </p><p>
 257     Google, Facebook, and other surveillance capitalists promise their customers
 258 (the advertisers) that if they use machine-learning tools trained on
 259 unimaginably large data sets of nonconsensually harvested personal
 260 information, they will be able to uncover ways to bypass the rational
 261 faculties of the public and direct their behavior, creating a stream of
 262 purchases, votes, and other desired outcomes.
 263   </p><div class="blockquote"><blockquote class="blockquote"><p>
 264       The impact of dominance far exceeds the impact of manipulation and should be
 265 central to our analysis and any remedies we seek.
 266     </p></blockquote></div><p>
 267     But there’s little evidence that this is happening. Instead, the predictions
 268 that surveillance capitalism delivers to its customers are much less
 269 impressive. Rather than finding ways to bypass our rational faculties,
 270 surveillance capitalists like Mark Zuckerberg mostly do one or more of three
 271 things:
 272   </p><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="segmenting"></a>1. Segmenting</h3></div></div></div><p>
 273       If you’re selling diapers, you have better luck if you pitch them to people
 274 in maternity wards. Not everyone who enters or leaves a maternity ward just
 275 had a baby, and not everyone who just had a baby is in the market for
 276 diapers. But having a baby is a really reliable correlate of being in the
 277 market for diapers, and being in a maternity ward is highly correlated with
 278 having a baby. Hence diaper ads around maternity wards (and even pitchmen
 279 for baby products, who haunt maternity wards with baskets full of freebies).
 280     </p><p>
 281       Le capitalisme de surveillance se segmente par milliard. Les vendeurs de
 282 couches peuvent aller bien au-delà des personnes dans les maternités (bien
 283 qu’ils puissent le faire aussi, avec des choses comme les annonces mobiles
 284 basées sur la localisation). Ils peuvent vous cibler selon que vous lisez
 285 des articles sur l’éducation des enfants, les couches ou une foule d’autres
 286 sujets, et l’exploration de données peut suggérer des mots-clés non évidents
 287 sur lesquels faire de la publicité. Ils peuvent vous cibler en fonction des
 288 articles que vous avez récemment lus. Ils peuvent vous cibler en fonction de
 289 ce que vous avez récemment acheté. Ils peuvent vous cibler selon que vous
 290 recevez des courriels ou des messages privés sur ces sujets – ou même si
 291 vous en parlez à haute voix (bien que Facebook et autres affirment de
 292 manière convaincante que cela ne se produit pas encore).
 293     </p><p>
 294       This is seriously creepy.
 295     </p><p>
 296       But it’s not mind control.
 297     </p><p>
 298       It doesn’t deprive you of your free will. It doesn’t trick you.
 299     </p><p>
 300       Think of how surveillance capitalism works in politics.  Surveillance
 301 capitalist companies sell political operatives the power to locate people
 302 who might be receptive to their pitch.  Candidates campaigning on finance
 303 industry corruption seek people struggling with debt; candidates campaigning
 304 on xenophobia seek out racists. Political operatives have always targeted
 305 their message whether their intentions were honorable or not: Union
 306 organizers set up pitches at factory gates, and white supremacists hand out
 307 fliers at John Birch Society meetings.
 308     </p><p>
 309       But this is an inexact and thus wasteful practice. The union organizer can’t
 310 know which worker to approach on the way out of the factory gates and may
 311 waste their time on a covert John Birch Society member; the white
 312 supremacist doesn’t know which of the Birchers are so delusional that making
 313 it to a meeting is as much as they can manage and which ones might be
 314 convinced to cross the country to carry a tiki torch through the streets of
 315 Charlottesville, Virginia.
 316     </p><p>
 317       Because targeting improves the yields on political pitches, it can
 318 accelerate the pace of political upheaval by making it possible for everyone
 319 who has secretly wished for the toppling of an autocrat — or just an 11-term
 320 incumbent politician — to find everyone else who feels the same way at very
 321 low cost. This has been critical to the rapid crystallization of recent
 322 political movements including Black Lives Matter and Occupy Wall Street as
 323 well as less savory players like the far-right white nationalist movements
 324 that marched in Charlottesville.
 325     </p><p>
 326       It’s important to differentiate this kind of political organizing from
 327 influence campaigns; finding people who secretly agree with you isn’t the
 328 same as convincing people to agree with you. The rise of phenomena like
 329 nonbinary or otherwise nonconforming gender identities is often
 330 characterized by reactionaries as the result of online brainwashing
 331 campaigns that convince impressionable people that they have been secretly
 332 queer all along.
 333     </p><p>
 334       But the personal accounts of those who have come out tell a different story
 335 where people who long harbored a secret about their gender were emboldened
 336 by others coming forward and where people who knew that they were different
 337 but lacked a vocabulary for discussing that difference learned the right
 338 words from these low-cost means of finding people and learning about their
 339 ideas.
 340     </p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="deception"></a>2. Deception</h3></div></div></div><p>
 341       Lies and fraud are pernicious, and surveillance capitalism supercharges them
 342 through targeting. If you want to sell a fraudulent payday loan or subprime
 343 mortgage, surveillance capitalism can help you find people who are both
 344 desperate and unsophisticated and thus receptive to your pitch. This
 345 accounts for the rise of many phenomena, like multilevel marketing schemes,
 346 in which deceptive claims about potential earnings and the efficacy of sales
 347 techniques are targeted at desperate people by advertising against search
 348 queries that indicate, for example, someone struggling with ill-advised
 349 loans.
 350     </p><p>
 351       Surveillance capitalism also abets fraud by making it easy to locate other
 352 people who have been similarly deceived, forming a community of people who
 353 reinforce one another’s false beliefs.  Think of <a class="ulink" href="https://www.vulture.com/2020/01/the-dream-podcast-review.html" target="_top">the
 354 forums</a> where people who are being victimized by multilevel marketing
 355 frauds gather to trade tips on how to improve their luck in peddling the
 356 product.
 357     </p><p>
 358       Sometimes, online deception involves replacing someone’s correct beliefs
 359 with incorrect ones, as it does in the anti-vaccination movement, whose
 360 victims are often people who start out believing in vaccines but are
 361 convinced by seemingly plausible evidence that leads them into the false
 362 belief that vaccines are harmful.
 363     </p><p>
 364       But it’s much more common for fraud to succeed when it doesn’t have to
 365 displace a true belief. When my daughter contracted head lice at daycare,
 366 one of the daycare workers told me I could get rid of them by treating her
 367 hair and scalp with olive oil. I didn’t know anything about head lice, and I
 368 assumed that the daycare worker did, so I tried it (it didn’t work, and it
 369 doesn’t work). It’s easy to end up with false beliefs when you simply don’t
 370 know any better and when those beliefs are conveyed by someone who seems to
 371 know what they’re doing.
 372     </p><p>
 373       This is pernicious and difficult — and it’s also the kind of thing the
 374 internet can help guard against by making true information available,
 375 especially in a form that exposes the underlying deliberations among parties
 376 with sharply divergent views, such as Wikipedia. But it’s not brainwashing;
 377 it’s fraud. In the <a class="ulink" href="https://datasociety.net/library/data-voids/" target="_top">majority of cases</a>,
 378 the victims of these fraud campaigns have an informational void filled in
 379 the customary way, by consulting a seemingly reliable source. If I look up
 380 the length of the Brooklyn Bridge and learn that it is 5,800 feet long, but
 381 in reality, it is 5,989 feet long, the underlying deception is a problem,
 382 but it’s a problem with a simple remedy. It’s a very different problem from
 383 the anti-vax issue in which someone’s true belief is displaced by a false
 384 one by means of sophisticated persuasion.
 385     </p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="domination"></a>3. Domination</h3></div></div></div><p>
 386       Surveillance capitalism is the result of monopoly. Monopoly is the cause,
 387 and surveillance capitalism and its negative outcomes are the effects of
 388 monopoly. I’ll get into this in depth later, but for now, suffice it to say
 389 that the tech industry has grown up with a radical theory of antitrust that
 390 has allowed companies to grow by merging with their rivals, buying up their
 391 nascent competitors, and expanding to control whole market verticals.
 392     </p><p>
 393       One example of how monopolism aids in persuasion is through dominance:
 394 Google makes editorial decisions about its algorithms that determine the
 395 sort order of the responses to our queries. If a cabal of fraudsters have
 396 set out to trick the world into thinking that the Brooklyn Bridge is 5,800
 397 feet long, and if Google gives a high search rank to this group in response
 398 to queries like <span class="quote">« <span class="quote">How long is the Brooklyn Bridge?</span> »</span> then the
 399 first eight or 10 screens’ worth of Google results could be wrong. And since
 400 most people don’t go beyond the first couple of results — let alone the
 401 first <span class="emphasis"><em>page</em></span> of results — Google’s choice means that many
 402 people will be deceived.
 403     </p><p>
 404       Google’s dominance over search — more than 86% of web searches are performed
 405 through Google — means that the way it orders its search results has an
 406 outsized effect on public beliefs. Ironically, Google claims this is why it
 407 can’t afford to have any transparency in its algorithm design: Google’s
 408 search dominance makes the results of its sorting too important to risk
 409 telling the world how it arrives at those results lest some bad actor
 410 discover a flaw in the ranking system and exploit it to push its point of
 411 view to the top of the search results. There’s an obvious remedy to a
 412 company that is too big to audit: break it up into smaller pieces.
 413     </p><p>
 414       Zuboff calls surveillance capitalism a <span class="quote">« <span class="quote">rogue capitalism</span> »</span> whose
 415 data-hoarding and machine-learning techniques rob us of our free will. But
 416 influence campaigns that seek to displace existing, correct beliefs with
 417 false ones have an effect that is small and temporary while monopolistic
 418 dominance over informational systems has massive, enduring
 419 effects. Controlling the results to the world’s search queries means
 420 controlling access both to arguments and their rebuttals and, thus, control
 421 over much of the world’s beliefs. If our concern is how corporations are
 422 foreclosing on our ability to make up our own minds and determine our own
 423 futures, the impact of dominance far exceeds the impact of manipulation and
 424 should be central to our analysis and any remedies we seek.
 425     </p></div><div class="sect2"><div class="titlepage"><div><div><h3 class="title"><a name="bypassing-our-rational-faculties"></a>4. Bypassing our rational faculties</h3></div></div></div><p>
 426       <span class="emphasis"><em>This</em></span> is the good stuff: using machine learning,
 427 <span class="quote">« <span class="quote">dark patterns,</span> »</span> engagement hacking, and other techniques to
 428 get us to do things that run counter to our better judgment. This is mind
 429 control.
 430     </p><p>
 431       Some of these techniques have proven devastatingly effective (if only in the
 432 short term). The use of countdown timers on a purchase completion page can
 433 create a sense of urgency that causes you to ignore the nagging internal
 434 voice suggesting that you should shop around or sleep on your decision. The
 435 use of people from your social graph in ads can provide <span class="quote">« <span class="quote">social
 436 proof</span> »</span> that a purchase is worth making. Even the auction system
 437 pioneered by eBay is calculated to play on our cognitive blind spots,
 438 letting us feel like we <span class="quote">« <span class="quote">own</span> »</span> something because we bid on it,
 439 thus encouraging us to bid again when we are outbid to ensure that
 440 <span class="quote">« <span class="quote">our</span> »</span> things stay ours.
 441     </p><p>
 442       Games are extraordinarily good at this. <span class="quote">« <span class="quote">Free to play</span> »</span> games
 443 manipulate us through many techniques, such as presenting players with a
 444 series of smoothly escalating challenges that create a sense of mastery and
 445 accomplishment but which sharply transition into a set of challenges that
 446 are impossible to overcome without paid upgrades. Add some social proof to
 447 the mix — a stream of notifications about how well your friends are faring —
 448 and before you know it, you’re buying virtual power-ups to get to the next
 449 level.
 450     </p><p>
 451       Companies have risen and fallen on these techniques, and the
 452 <span class="quote">« <span class="quote">fallen</span> »</span> part is worth paying attention to. In general, living
 453 things adapt to stimulus: Something that is very compelling or noteworthy
 454 when you first encounter it fades with repetition until you stop noticing it
 455 altogether. Consider the refrigerator hum that irritates you when it starts
 456 up but disappears into the background so thoroughly that you only notice it
 457 when it stops again.
 458     </p><p>
 459       That’s why behavioral conditioning uses <span class="quote">« <span class="quote">intermittent reinforcement
 460 schedules.</span> »</span> Instead of giving you a steady drip of encouragement or
 461 setbacks, games and gamified services scatter rewards on a randomized
 462 schedule — often enough to keep you interested and random enough that you
 463 can never quite find the pattern that would make it boring.
 464     </p><p>
 465       Intermittent reinforcement is a powerful behavioral tool, but it also
 466 represents a collective action problem for surveillance capitalism. The
 467 <span class="quote">« <span class="quote">engagement techniques</span> »</span> invented by the behaviorists of
 468 surveillance capitalist companies are quickly copied across the whole sector
 469 so that what starts as a mysteriously compelling fillip in the design of a
 470 service—like <span class="quote">« <span class="quote">pull to refresh</span> »</span> or alerts when someone likes
 471 your posts or side quests that your characters get invited to while in the
 472 midst of main quests—quickly becomes dully ubiquitous. The
 473 impossible-to-nail-down nonpattern of randomized drips from your phone
 474 becomes a grey-noise wall of sound as every single app and site starts to
 475 make use of whatever seems to be working at the time.
 476     </p><p>
 477       From the surveillance capitalist’s point of view, our adaptive capacity is
 478 like a harmful bacterium that deprives it of its food source — our attention
 479 — and novel techniques for snagging that attention are like new antibiotics
 480 that can be used to breach our defenses and destroy our
 481 self-determination. And there <span class="emphasis"><em>are</em></span> techniques like
 482 that. Who can forget the Great Zynga Epidemic, when all of our friends were
 483 caught in <span class="emphasis"><em>FarmVille</em></span>’s endless, mindless dopamine loops?
 484 But every new attention-commanding technique is jumped on by the whole
 485 industry and used so indiscriminately that antibiotic resistance sets
 486 in. Given enough repetition, almost all of us develop immunity to even the
 487 most powerful techniques — by 2013, two years after Zynga’s peak, its user
 488 base had halved.
 489     </p><p>
 490       Not everyone, of course. Some people never adapt to stimulus, just as some
 491 people never stop hearing the hum of the refrigerator.  This is why most
 492 people who are exposed to slot machines play them for a while and then move
 493 on while a small and tragic minority liquidate their kids’ college funds,
 494 buy adult diapers, and position themselves in front of a machine until they
 495 collapse.
 496     </p><p>
 497       But surveillance capitalism’s margins on behavioral modification
 498 suck. Tripling the rate at which someone buys a widget sounds great <a class="ulink" href="https://www.forbes.com/sites/priceonomics/2018/03/09/the-advertising-conversion-rates-for-every-major-tech-platform/#2f6a67485957" target="_top">unless
 499 the base rate is way less than 1%</a> with an improved rate of… still
 500 less than 1%. Even penny slot machines pull down pennies for every spin
 501 while surveillance capitalism rakes in infinitesimal penny fractions.
 502     </p><p>
 503       Slot machines’ high returns mean that they can be profitable just by
 504 draining the fortunes of the small rump of people who are pathologically
 505 vulnerable to them and unable to adapt to their tricks. But surveillance
 506 capitalism can’t survive on the fractional pennies it brings down from that
 507 vulnerable sliver — that’s why, after the Great Zynga Epidemic had finally
 508 burned itself out, the small number of still-addicted players left behind
 509 couldn’t sustain it as a global phenomenon. And new powerful attention
 510 weapons aren’t easy to find, as is evidenced by the long years since the
 511 last time Zynga had a hit. Despite the hundreds of millions of dollars that
 512 Zynga has to spend on developing new tools to blast through our adaptation,
 513 it has never managed to repeat the lucky accident that let it snag so much
 514 of our attention for a brief moment in 2009. Powerhouses like Supercell have
 515 fared a little better, but they are rare and throw away many failures for
 516 every success.
 517     </p><p>
 518       The vulnerability of small segments of the population to dramatic, efficient
 519 corporate manipulation is a real concern that’s worthy of our attention and
 520 energy. But it’s not an existential threat to society.
 521     </p></div></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="if-data-is-the-new-oil-then-surveillance-capitalisms-engine-has-a-leak"></a>If data is the new oil, then surveillance capitalism’s engine has a leak</h2></div></div></div><p>
 522     This adaptation problem offers an explanation for one of surveillance
 523 capitalism’s most alarming traits: its relentless hunger for data and its
 524 endless expansion of data-gathering capabilities through the spread of
 525 sensors, online surveillance, and acquisition of data streams from third
 526 parties.
 527   </p><p>
 528     Zuboff observes this phenomenon and concludes that data must be very
 529 valuable if surveillance capitalism is so hungry for it. (In her words:
 530 <span class="quote">« <span class="quote">Just as industrial capitalism was driven to the continuous
 531 intensification of the means of production, so surveillance capitalists and
 532 their market players are now locked into the continuous intensification of
 533 the means of behavioral modification and the gathering might of
 534 instrumentarian power.</span> »</span>) But what if the voracious appetite is
 535 because data has such a short half-life — because people become inured so
 536 quickly to new, data-driven persuasion techniques — that the companies are
 537 locked in an arms race with our limbic system? What if it’s all a Red
 538 Queen’s race where they have to run ever faster — collect ever-more data —
 539 just to stay in the same spot?
 540   </p><p>
 541     Of course, all of Big Tech’s persuasion techniques work in concert with one
 542 another, and collecting data is useful beyond mere behavioral trickery.
 543   </p><p>
 544     If someone wants to recruit you to buy a refrigerator or join a pogrom, they
 545 might use profiling and targeting to send messages to people they judge to
 546 be good sales prospects. The messages themselves may be deceptive, making
 547 claims about things you’re not very knowledgeable about (food safety and
 548 energy efficiency or eugenics and historical claims about racial
 549 superiority). They might use search engine optimization and/or armies of
 550 fake reviewers and commenters and/or paid placement to dominate the
 551 discourse so that any search for further information takes you back to their
 552 messages.  And finally, they may refine the different pitches using machine
 553 learning and other techniques to figure out what kind of pitch works best on
 554 someone like you.
 555   </p><p>
 556     Each phase of this process benefits from surveillance: The more data they
 557 have, the more precisely they can profile you and target you with specific
 558 messages. Think of how you’d sell a fridge if you knew that the warranty on
 559 your prospect’s fridge just expired and that they were expecting a tax
 560 rebate in April.
 561   </p><p>
 562     Also, the more data they have, the better they can craft deceptive messages
 563 — if I know that you’re into genealogy, I might not try to feed you
 564 pseudoscience about genetic differences between <span class="quote">« <span class="quote">races,</span> »</span>
 565 sticking instead to conspiratorial secret histories of <span class="quote">« <span class="quote">demographic
 566 replacement</span> »</span> and the like.
 567   </p><p>
 568     Facebook also helps you locate people who have the same odious or antisocial
 569 views as you. It makes it possible to find other people who want to carry
 570 tiki torches through the streets of Charlottesville in Confederate
 571 cosplay. It can help you find other people who want to join your militia and
 572 go to the border to look for undocumented migrants to terrorize. It can help
 573 you find people who share your belief that vaccines are poison and that the
 574 Earth is flat.
 575   </p><p>
 576     There is one way in which targeted advertising uniquely benefits those
 577 advocating for socially unacceptable causes: It is invisible.  Racism is
 578 widely geographically dispersed, and there are few places where racists —
 579 and only racists — gather. This is similar to the problem of selling
 580 refrigerators in that potential refrigerator purchasers are geographically
 581 dispersed and there are few places where you can buy an ad that will be
 582 primarily seen by refrigerator customers. But buying a refrigerator is
 583 socially acceptable while being a Nazi is not, so you can buy a billboard or
 584 advertise in the newspaper sports section for your refrigerator business,
 585 and the only potential downside is that your ad will be seen by a lot of
 586 people who don’t want refrigerators, resulting in a lot of wasted expense.
 587   </p><p>
 588     But even if you wanted to advertise your Nazi movement on a billboard or
 589 prime-time TV or the sports section, you would struggle to find anyone
 590 willing to sell you the space for your ad partly because they disagree with
 591 your views and partly because they fear censure (boycott, reputational
 592 damage, etc.) from other people who disagree with your views.
 593   </p><p>
 594     Targeted ads solve this problem: On the internet, every ad unit can be
 595 different for every person, meaning that you can buy ads that are only shown
 596 to people who appear to be Nazis and not to people who hate Nazis. When
 597 there’s spillover — when someone who hates racism is shown a racist
 598 recruiting ad — there is some fallout; the platform or publication might get
 599 an angry public or private denunciation. But the nature of the risk assumed
 600 by an online ad buyer is different than the risks to a traditional publisher
 601 or billboard owner who might want to run a Nazi ad.
 602   </p><p>
 603     Online ads are placed by algorithms that broker between a diverse ecosystem
 604 of self-serve ad platforms that anyone can buy an ad through, so the Nazi ad
 605 that slips onto your favorite online publication isn’t seen as their moral
 606 failing but rather as a failure in some distant, upstream ad supplier. When
 607 a publication gets a complaint about an offensive ad that’s appearing in one
 608 of its units, it can take some steps to block that ad, but the Nazi might
 609 buy a slightly different ad from a different broker serving the same
 610 unit. And in any event, internet users increasingly understand that when
 611 they see an ad, it’s likely that the advertiser did not choose that
 612 publication and that the publication has no idea who its advertisers are.
 613   </p><p>
 614     These layers of indirection between advertisers and publishers serve as
 615 moral buffers: Today’s moral consensus is largely that publishers shouldn’t
 616 be held responsible for the ads that appear on their pages because they’re
 617 not actively choosing to put those ads there.  Because of this, Nazis are
 618 able to overcome significant barriers to organizing their movement.
 619   </p><p>
 620     Data has a complex relationship with domination. Being able to spy on your
 621 customers can alert you to their preferences for your rivals and allow you
 622 to head off your rivals at the pass.
 623   </p><p>
 624     More importantly, if you can dominate the information space while also
 625 gathering data, then you make other deceptive tactics stronger because it’s
 626 harder to break out of the web of deceit you’re spinning. Domination — that
 627 is, ultimately becoming a monopoly — and not the data itself is the
 628 supercharger that makes every tactic worth pursuing because monopolistic
 629 domination deprives your target of an escape route.
 630   </p><p>
 631     If you’re a Nazi who wants to ensure that your prospects primarily see
 632 deceptive, confirming information when they search for more, you can improve
 633 your odds by seeding the search terms they use through your initial
 634 communications. You don’t need to own the top 10 results for <span class="quote">« <span class="quote">voter
 635 suppression</span> »</span> if you can convince your marks to confine their search
 636 terms to <span class="quote">« <span class="quote">voter fraud,</span> »</span> which throws up a very different set of
 637 search results.
 638   </p><p>
 639     Surveillance capitalists are like stage mentalists who claim that their
 640 extraordinary insights into human behavior let them guess the word that you
 641 wrote down and folded up in your pocket but who really use shills, hidden
 642 cameras, sleight of hand, and brute-force memorization to amaze you.
 643   </p><p>
 644     Or perhaps they’re more like pick-up artists, the misogynistic cult that
 645 promises to help awkward men have sex with women by teaching them
 646 <span class="quote">« <span class="quote">neurolinguistic programming</span> »</span> phrases, body language
 647 techniques, and psychological manipulation tactics like
 648 <span class="quote">« <span class="quote">negging</span> »</span> — offering unsolicited negative feedback to women to
 649 lower their self-esteem and prick their interest.
 650   </p><p>
 651     Some pick-up artists eventually manage to convince women to go home with
 652 them, but it’s not because these men have figured out how to bypass women’s
 653 critical faculties. Rather, pick-up artists’ <span class="quote">« <span class="quote">success</span> »</span> stories
 654 are a mix of women who were incapable of giving consent, women who were
 655 coerced, women who were intoxicated, self-destructive women, and a few women
 656 who were sober and in command of their faculties but who didn’t realize
 657 straightaway that they were with terrible men but rectified the error as
 658 soon as they could.
 659   </p><p>
 660     Pick-up artists <span class="emphasis"><em>believe</em></span> they have figured out a secret
 661 back door that bypasses women’s critical faculties, but they haven’t. Many
 662 of the tactics they deploy, like negging, became the butt of jokes (just
 663 like people joke about bad ad targeting), and there’s a good chance that
 664 anyone they try these tactics on will immediately recognize them and dismiss
 665 the men who use them as irredeemable losers.
 666   </p><p>
 667     Pick-up artists are proof that people can believe they have developed a
 668 system of mind control <span class="emphasis"><em>even when it doesn’t
 669 work</em></span>. Pick-up artists simply exploit the fact that
 670 one-in-a-million chances can come through for you if you make a million
 671 attempts, and then they assume that the other 999,999 times, they simply
 672 performed the technique incorrectly and commit themselves to doing better
 673 next time. There’s only one group of people who find pick-up artist lore
 674 reliably convincing: other would-be pick-up artists whose anxiety and
 675 insecurity make them vulnerable to scammers and delusional men who convince
 676 them that if they pay for tutelage and follow instructions, then they will
 677 someday succeed. Pick-up artists assume they fail to entice women because
 678 they are bad at being pick-up artists, not because pick-up artistry is
 679 bullshit. Pick-up artists are bad at selling themselves to women, but
 680 they’re much better at selling themselves to men who pay to learn the
 681 secrets of pick-up artistry.
 682   </p><p>
 683     Department store pioneer John Wanamaker is said to have lamented,
 684 <span class="quote">« <span class="quote">Half the money I spend on advertising is wasted; the trouble is I
 685 don’t know which half.</span> »</span> The fact that Wanamaker thought that only
 686 half of his advertising spending was wasted is a tribute to the
 687 persuasiveness of advertising executives, who are <span class="emphasis"><em>much</em></span>
 688 better at convincing potential clients to buy their services than they are
 689 at convincing the general public to buy their clients’ wares.
 690   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="what-is-facebook"></a>What is Facebook?</h2></div></div></div><p>
 691     Facebook is heralded as the origin of all of our modern plagues, and it’s
 692 not hard to see why. Some tech companies want to lock their users in but
 693 make their money by monopolizing access to the market for apps for their
 694 devices and gouging them on prices rather than by spying on them (like
 695 Apple). Some companies don’t care about locking in users because they’ve
 696 figured out how to spy on them no matter where they are and what they’re
 697 doing and can turn that surveillance into money (Google). Facebook alone
 698 among the Western tech giants has built a business based on locking in its
 699 users <span class="emphasis"><em>and</em></span> spying on them all the time.
 700   </p><p>
 701     Facebook’s surveillance regime is really without parallel in the Western
 702 world. Though Facebook tries to prevent itself from being visible on the
 703 public web, hiding most of what goes on there from people unless they’re
 704 logged into Facebook, the company has nevertheless booby-trapped the entire
 705 web with surveillance tools in the form of Facebook <span class="quote">« <span class="quote">Like</span> »</span>
 706 buttons that web publishers include on their sites to boost their Facebook
 707 profiles. Facebook also makes various libraries and other useful code
 708 snippets available to web publishers that act as surveillance tendrils on
 709 the sites where they’re used, funneling information about visitors to the
 710 site — newspapers, dating sites, message boards — to Facebook.
 711   </p><div class="blockquote"><blockquote class="blockquote"><p>
 712       Big Tech is able to practice surveillance not just because it is tech but
 713 because it is <span class="emphasis"><em>big</em></span>.
 714     </p></blockquote></div><p>
 715     Facebook offers similar tools to app developers, so the apps — games, fart
 716 machines, business review services, apps for keeping abreast of your kid’s
 717 schooling — you use will send information about your activities to Facebook
 718 even if you don’t have a Facebook account and even if you don’t download or
 719 use Facebook apps. On top of all that, Facebook buys data from third-party
 720 brokers on shopping habits, physical location, use of <span class="quote">« <span class="quote">loyalty</span> »</span>
 721 programs, financial transactions, etc., and cross-references that with the
 722 dossiers it develops on activity on Facebook and with apps and the public
 723 web.
 724   </p><p>
 725     Though it’s easy to integrate the web with Facebook — linking to news
 726 stories and such — Facebook products are generally not available to be
 727 integrated back into the web itself. You can embed a tweet in a Facebook
 728 post, but if you embed a Facebook post in a tweet, you just get a link back
 729 to Facebook and must log in before you can see it. Facebook has used extreme
 730 technological and legal countermeasures to prevent rivals from allowing
 731 their users to embed Facebook snippets in competing services or to create
 732 alternative interfaces to Facebook that merge your Facebook inbox with those
 733 of other services that you use.
 734   </p><p>
 735     And Facebook is incredibly popular, with 2.3 billion claimed users (though
 736 many believe this figure to be inflated). Facebook has been used to organize
 737 genocidal pogroms, racist riots, anti-vaccination movements, flat Earth
 738 cults, and the political lives of some of the world’s ugliest, most brutal
 739 autocrats. There are some really alarming things going on in the world, and
 740 Facebook is implicated in many of them, so it’s easy to conclude that these
 741 bad things are the result of Facebook’s mind-control system, which it rents
 742 out to anyone with a few bucks to spend.
 743   </p><p>
 744     To understand what role Facebook plays in the formulation and mobilization
 745 of antisocial movements, we need to understand the dual nature of Facebook.
 746   </p><p>
 747     Because it has a lot of users and a lot of data about those users, Facebook
 748 is a very efficient tool for locating people with hard-to-find traits, the
 749 kinds of traits that are widely diffused in the population such that
 750 advertisers have historically struggled to find a cost-effective way to
 751 reach them. Think back to refrigerators: Most of us only replace our major
 752 appliances a few times in our entire lives. If you’re a refrigerator
 753 manufacturer or retailer, you have these brief windows in the life of a
 754 consumer during which they are pondering a purchase, and you have to somehow
 755 reach them. Anyone who’s ever registered a title change after buying a house
 756 can attest that appliance manufacturers are incredibly desperate to reach
 757 anyone who has even the slenderest chance of being in the market for a new
 758 fridge.
 759   </p><p>
 760     Facebook makes finding people shopping for refrigerators a
 761 <span class="emphasis"><em>lot</em></span> easier. It can target ads to people who’ve
 762 registered a new home purchase, to people who’ve searched for refrigerator
 763 buying advice, to people who have complained about their fridge dying, or
 764 any combination thereof. It can even target people who’ve recently bought
 765 <span class="emphasis"><em>other</em></span> kitchen appliances on the theory that someone
 766 who’s just replaced their stove and dishwasher might be in a fridge-buying
 767 kind of mood. The vast majority of people who are reached by these ads will
 768 not be in the market for a new fridge, but — crucially — the percentage of
 769 people who <span class="emphasis"><em>are</em></span> looking for fridges that these ads reach
 770 is <span class="emphasis"><em>much</em></span> larger than it is than for any group that might
 771 be subjected to traditional, offline targeted refrigerator marketing.
 772   </p><p>
 773     Facebook also makes it a lot easier to find people who have the same rare
 774 disease as you, which might have been impossible in earlier eras — the
 775 closest fellow sufferer might otherwise be hundreds of miles away. It makes
 776 it easier to find people who went to the same high school as you even though
 777 decades have passed and your former classmates have all been scattered to
 778 the four corners of the Earth.
 779   </p><p>
 780     Facebook also makes it much easier to find people who hold the same rare
 781 political beliefs as you. If you’ve always harbored a secret affinity for
 782 socialism but never dared utter this aloud lest you be demonized by your
 783 neighbors, Facebook can help you discover other people who feel the same way
 784 (and it might just demonstrate to you that your affinity is more widespread
 785 than you ever suspected). It can make it easier to find people who share
 786 your sexual identity.  And again, it can help you to understand that what
 787 you thought was a shameful secret that affected only you was really a widely
 788 shared trait, giving you both comfort and the courage to come out to the
 789 people in your life.
 790   </p><p>
 791     All of this presents a dilemma for Facebook: Targeting makes the company’s
 792 ads more effective than traditional ads, but it also lets advertisers see
 793 just how effective their ads are. While advertisers are pleased to learn
 794 that Facebook ads are more effective than ads on systems with less
 795 sophisticated targeting, advertisers can also see that in nearly every case,
 796 the people who see their ads ignore them. Or, at best, the ads work on a
 797 subconscious level, creating nebulous unmeasurables like <span class="quote">« <span class="quote">brand
 798 recognition.</span> »</span> This means that the price per ad is very low in nearly
 799 every case.
 800   </p><p>
 801     To make things worse, many Facebook groups spark precious little
 802 discussion. Your little-league soccer team, the people with the same rare
 803 disease as you, and the people you share a political affinity with may
 804 exchange the odd flurry of messages at critical junctures, but on a daily
 805 basis, there’s not much to say to your old high school chums or other
 806 hockey-card collectors.
 807   </p><p>
 808     With nothing but <span class="quote">« <span class="quote">organic</span> »</span> discussion, Facebook would not
 809 generate enough traffic to sell enough ads to make the money it needs to
 810 continually expand by buying up its competitors while returning handsome
 811 sums to its investors.
 812   </p><p>
 813     So Facebook has to gin up traffic by sidetracking its own forums: Every time
 814 Facebook’s algorithm injects controversial materials — inflammatory
 815 political articles, conspiracy theories, outrage stories — into a group, it
 816 can hijack that group’s nominal purpose with its desultory discussions and
 817 supercharge those discussions by turning them into bitter, unproductive
 818 arguments that drag on and on. Facebook is optimized for engagement, not
 819 happiness, and it turns out that automated systems are pretty good at
 820 figuring out things that people will get angry about.
 821   </p><p>
 822     Facebook <span class="emphasis"><em>can</em></span> modify our behavior but only in a couple
 823 of trivial ways. First, it can lock in all your friends and family members
 824 so that you check and check and check with Facebook to find out what they
 825 are up to; and second, it can make you angry and anxious. It can force you
 826 to choose between being interrupted constantly by updates — a process that
 827 breaks your concentration and makes it hard to be introspective — and
 828 staying in touch with your friends. This is a very limited form of mind
 829 control, and it can only really make us miserable, angry, and anxious.
 830   </p><p>
 831     This is why Facebook’s targeting systems — both the ones it shows to
 832 advertisers and the ones that let users find people who share their
 833 interests — are so next-gen and smooth and easy to use as well as why its
 834 message boards have a toolset that seems like it hasn’t changed since the
 835 mid-2000s. If Facebook delivered an equally flexible, sophisticated
 836 message-reading system to its users, those users could defend themselves
 837 against being nonconsensually eyeball-fucked with Donald Trump headlines.
 838   </p><p>
 839     The more time you spend on Facebook, the more ads it gets to show you. The
 840 solution to Facebook’s ads only working one in a thousand times is for the
 841 company to try to increase how much time you spend on Facebook by a factor
 842 of a thousand. Rather than thinking of Facebook as a company that has
 843 figured out how to show you exactly the right ad in exactly the right way to
 844 get you to do what its advertisers want, think of it as a company that has
 845 figured out how to make you slog through an endless torrent of arguments
 846 even though they make you miserable, spending so much time on the site that
 847 it eventually shows you at least one ad that you respond to.
 848   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="monopoly-and-the-right-to-the-future-tense"></a>Monopoly and the right to the future tense</h2></div></div></div><p>
 849     Zuboff and her cohort are particularly alarmed at the extent to which
 850 surveillance allows corporations to influence our decisions, taking away
 851 something she poetically calls <span class="quote">« <span class="quote">the right to the future tense</span> »</span>
 852 — that is, the right to decide for yourself what you will do in the future.
 853   </p><p>
 854     It’s true that advertising can tip the scales one way or another: When
 855 you’re thinking of buying a fridge, a timely fridge ad might end the search
 856 on the spot. But Zuboff puts enormous and undue weight on the persuasive
 857 power of surveillance-based influence techniques. Most of these don’t work
 858 very well, and the ones that do won’t work for very long. The makers of
 859 these influence tools are confident they will someday refine them into
 860 systems of total control, but they are hardly unbiased observers, and the
 861 risks from their dreams coming true are very speculative.
 862   </p><p>
 863     By contrast, Zuboff is rather sanguine about 40 years of lax antitrust
 864 practice that has allowed a handful of companies to dominate the internet,
 865 ushering in an information age with, <a class="ulink" href="https://twitter.com/tveastman/status/1069674780826071040" target="_top">as one person
 866 on Twitter noted</a>, five giant websites each filled with screenshots
 867 of the other four.
 868   </p><p>
 869     However, if we are to be alarmed that we might lose the right to choose for
 870 ourselves what our future will hold, then monopoly’s nonspeculative,
 871 concrete, here-and-now harms should be front and center in our debate over
 872 tech policy.
 873   </p><p>
 874     Start with <span class="quote">« <span class="quote">digital rights management.</span> »</span> In 1998, Bill Clinton
 875 signed the Digital Millennium Copyright Act (DMCA) into law. It’s a complex
 876 piece of legislation with many controversial clauses but none more so than
 877 Section 1201, the <span class="quote">« <span class="quote">anti-circumvention</span> »</span> rule.
 878   </p><p>
 879     This is a blanket ban on tampering with systems that restrict access to
 880 copyrighted works. The ban is so thoroughgoing that it prohibits removing a
 881 copyright lock even when no copyright infringement takes place. This is by
 882 design: The activities that the DMCA’s Section 1201 sets out to ban are not
 883 copyright infringements; rather, they are legal activities that frustrate
 884 manufacturers’ commercial plans.
 885   </p><p>
 886     For example, Section 1201’s first major application was on DVD players as a
 887 means of enforcing the region coding built into those devices. DVD-CCA, the
 888 body that standardized DVDs and DVD players, divided the world into six
 889 regions and specified that DVD players must check each disc to determine
 890 which regions it was authorized to be played in. DVD players would have
 891 their own corresponding region (a DVD player bought in the U.S. would be
 892 region 1 while one bought in India would be region 5). If the player and the
 893 disc’s region matched, the player would play the disc; otherwise, it would
 894 reject it.
 895   </p><p>
 896     However, watching a lawfully produced disc in a country other than the one
 897 where you purchased it is not copyright infringement — it’s the
 898 opposite. Copyright law imposes this duty on customers for a movie: You must
 899 go into a store, find a licensed disc, and pay the asking price. Do that —
 900 and <span class="emphasis"><em>nothing else</em></span> — and you and copyright are square
 901 with one another.
 902   </p><p>
 903     The fact that a movie studio wants to charge Indians less than Americans or
 904 release in Australia later than it releases in the U.K.  has no bearing on
 905 copyright law. Once you lawfully acquire a DVD, it is no copyright
 906 infringement to watch it no matter where you happen to be.
 907   </p><p>
 908     So DVD and DVD player manufacturers would not be able to use accusations of
 909 abetting copyright infringement to punish manufacturers who made
 910 noncompliant players that would play discs from any region or repair shops
 911 that modified players to let you watch out-of-region discs or software
 912 programmers who created programs to let you do this.
 913   </p><p>
 914     That’s where Section 1201 of the DMCA comes in: By banning tampering with an
 915 <span class="quote">« <span class="quote">access control,</span> »</span> the rule gave manufacturers and rights
 916 holders standing to sue competitors who released superior products with
 917 lawful features that the market demanded (in this case, region-free
 918 players).
 919   </p><p>
 920     This is an odious scam against consumers, but as time went by, Section 1201
 921 grew to encompass a rapidly expanding constellation of devices and services
 922 as canny manufacturers have realized certain things:
 923   </p><div class="itemizedlist"><ul class="itemizedlist compact" style="list-style-type: disc; "><li class="listitem"><p>
 924         Any device with software in it contains a <span class="quote">« <span class="quote">copyrighted work</span> »</span> —
 925 i.e., the software.
 926       </p></li><li class="listitem"><p>
 927         A device can be designed so that reconfiguring the software requires
 928 bypassing an <span class="quote">« <span class="quote">access control for copyrighted works,</span> »</span> which is a
 929 potential felony under Section 1201.
 930       </p></li><li class="listitem"><p>
 931         Thus, companies can control their customers’ behavior after they take home
 932 their purchases by designing products so that all unpermitted uses require
 933 modifications that fall afoul of Section 1201.
 934       </p></li></ul></div><p>
 935     Section 1201 then becomes a means for manufacturers of all descriptions to
 936 force their customers to arrange their affairs to benefit the manufacturers’
 937 shareholders instead of themselves.
 938   </p><p>
 939     This manifests in many ways: from a new generation of inkjet printers that
 940 use countermeasures to prevent third-party ink that cannot be bypassed
 941 without legal risks to similar systems in tractors that prevent third-party
 942 technicians from swapping in the manufacturer’s own parts that are not
 943 recognized by the tractor’s control system until it is supplied with a
 944 manufacturer’s unlock code.
 945   </p><p>
 946     Closer to home, Apple’s iPhones use these measures to prevent both
 947 third-party service and third-party software installation. This allows Apple
 948 to decide when an iPhone is beyond repair and must be shredded and
 949 landfilled as opposed to the iPhone’s purchaser. (Apple is notorious for its
 950 environmentally catastrophic policy of destroying old electronics rather
 951 than permitting them to be cannibalized for parts.) This is a very useful
 952 power to wield, especially in light of CEO Tim Cook’s January 2019 warning
 953 to investors that the company’s profits are endangered by customers choosing
 954 to hold onto their phones for longer rather than replacing them.
 955   </p><p>
 956     Apple’s use of copyright locks also allows it to establish a monopoly over
 957 how its customers acquire software for their mobile devices. The App Store’s
 958 commercial terms guarantee Apple a share of all revenues generated by the
 959 apps sold there, meaning that Apple gets paid when you buy an app from its
 960 store and then continues to get paid every time you buy something using that
 961 app. This comes out of the bottom line of software developers, who must
 962 either charge more or accept lower profits for their products.
 963   </p><p>
 964     Crucially, Apple’s use of copyright locks gives it the power to make
 965 editorial decisions about which apps you may and may not install on your own
 966 device. Apple has used this power to <a class="ulink" href="https://www.telegraph.co.uk/technology/apple/5982243/Apple-bans-dictionary-from-App-Store-over-swear-words.html" target="_top">reject
 967 dictionaries</a> for containing obscene words; to <a class="ulink" href="https://www.vice.com/en_us/article/538kan/apple-just-banned-the-app-that-tracks-us-drone-strikes-again" target="_top">limit
 968 political speech</a>, especially from apps that make sensitive political
 969 commentary such as an app that notifies you every time a U.S. drone kills
 970 someone somewhere in the world; and to <a class="ulink" href="https://www.eurogamer.net/articles/2016-05-19-palestinian-indie-game-must-not-be-called-a-game-apple-says" target="_top">object
 971 to a game</a> that commented on the Israel-Palestine conflict.
 972   </p><p>
 973     Apple often justifies monopoly power over software installation in the name
 974 of security, arguing that its vetting of apps for its store means that it
 975 can guard its users against apps that contain surveillance code. But this
 976 cuts both ways. In China, the government <a class="ulink" href="https://www.ft.com/content/ad42e536-cf36-11e7-b781-794ce08b24dc" target="_top">ordered
 977 Apple to prohibit the sale of privacy tools</a> like VPNs with the
 978 exception of VPNs that had deliberately introduced flaws designed to let the
 979 Chinese state eavesdrop on users. Because Apple uses technological
 980 countermeasures — with legal backstops — to block customers from installing
 981 unauthorized apps, Chinese iPhone owners cannot readily (or legally) acquire
 982 VPNs that would protect them from Chinese state snooping.
 983   </p><p>
 984     Zuboff calls surveillance capitalism a <span class="quote">« <span class="quote">rogue capitalism.</span> »</span>
 985 Theoreticians of capitalism claim that its virtue is that it <a class="ulink" href="https://en.wikipedia.org/wiki/Price_signal" target="_top">aggregates information in
 986 the form of consumers’ decisions</a>, producing efficient
 987 markets. Surveillance capitalism’s supposed power to rob its victims of
 988 their free will through computationally supercharged influence campaigns
 989 means that our markets no longer aggregate customers’ decisions because we
 990 customers no longer decide — we are given orders by surveillance
 991 capitalism’s mind-control rays.
 992   </p><p>
 993     If our concern is that markets cease to function when consumers can no
 994 longer make choices, then copyright locks should concern us at
 995 <span class="emphasis"><em>least</em></span> as much as influence campaigns. An influence
 996 campaign might nudge you to buy a certain brand of phone; but the copyright
 997 locks on that phone absolutely determine where you get it serviced, which
 998 apps can run on it, and when you have to throw it away rather than fixing
 999 it.
1000   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="search-order-and-the-right-to-the-future-tense"></a>Search order and the right to the future tense</h2></div></div></div><p>
1001     Markets are posed as a kind of magic: By discovering otherwise hidden
1002 information conveyed by the free choices of consumers, those consumers’
1003 local knowledge is integrated into a self-correcting system that makes
1004 efficient allocations—more efficient than any computer could calculate. But
1005 monopolies are incompatible with that notion. When you only have one app
1006 store, the owner of the store — not the consumer — decides on the range of
1007 choices. As Boss Tweed once said, <span class="quote">« <span class="quote">I don’t care who does the electing,
1008 so long as I get to do the nominating.</span> »</span> A monopolized market is an
1009 election whose candidates are chosen by the monopolist.
1010   </p><p>
1011     This ballot rigging is made more pernicious by the existence of monopolies
1012 over search order. Google’s search market share is about 90%. When Google’s
1013 ranking algorithm puts a result for a popular search term in its top 10,
1014 that helps determine the behavior of millions of people. If Google’s answer
1015 to <span class="quote">« <span class="quote">Are vaccines dangerous?</span> »</span> is a page that rebuts anti-vax
1016 conspiracy theories, then a sizable portion of the public will learn that
1017 vaccines are safe. If, on the other hand, Google sends those people to a
1018 site affirming the anti-vax conspiracies, a sizable portion of those
1019 millions will come away convinced that vaccines are dangerous.
1020   </p><p>
1021     Google’s algorithm is often tricked into serving disinformation as a
1022 prominent search result. But in these cases, Google isn’t persuading people
1023 to change their minds; it’s just presenting something untrue as fact when
1024 the user has no cause to doubt it.
1025   </p><p>
1026     This is true whether the search is for <span class="quote">« <span class="quote">Are vaccines
1027 dangerous?</span> »</span> or <span class="quote">« <span class="quote">best restaurants near me.</span> »</span> Most users
1028 will never look past the first page of search results, and when the
1029 overwhelming majority of people all use the same search engine, the ranking
1030 algorithm deployed by that search engine will determine myriad outcomes
1031 (whether to adopt a child, whether to have cancer surgery, where to eat
1032 dinner, where to move, where to apply for a job) to a degree that vastly
1033 outstrips any behavioral outcomes dictated by algorithmic persuasion
1034 techniques.
1035   </p><p>
1036     Many of the questions we ask search engines have no empirically correct
1037 answers: <span class="quote">« <span class="quote">Where should I eat dinner?</span> »</span> is not an objective
1038 question. Even questions that do have correct answers (<span class="quote">« <span class="quote">Are vaccines
1039 dangerous?</span> »</span>) don’t have one empirically superior source for that
1040 answer. Many pages affirm the safety of vaccines, so which one goes first?
1041 Under conditions of competition, consumers can choose from many search
1042 engines and stick with the one whose algorithmic judgment suits them best,
1043 but under conditions of monopoly, we all get our answers from the same
1044 place.
1045   </p><p>
1046     Google’s search dominance isn’t a matter of pure merit: The company has
1047 leveraged many tactics that would have been prohibited under classical,
1048 pre-Ronald-Reagan antitrust enforcement standards to attain its
1049 dominance. After all, this is a company that has developed two major
1050 products: a really good search engine and a pretty good Hotmail clone. Every
1051 other major success it’s had — Android, YouTube, Google Maps, etc. — has
1052 come through an acquisition of a nascent competitor. Many of the company’s
1053 key divisions, such as the advertising technology of DoubleClick, violate
1054 the historical antitrust principle of structural separation, which forbade
1055 firms from owning subsidiaries that competed with their
1056 customers. Railroads, for example, were barred from owning freight companies
1057 that competed with the shippers whose freight they carried.
1058   </p><p>
1059     If we’re worried about giant companies subverting markets by stripping
1060 consumers of their ability to make free choices, then vigorous antitrust
1061 enforcement seems like an excellent remedy. If we’d denied Google the right
1062 to effect its many mergers, we would also have probably denied it its total
1063 search dominance. Without that dominance, the pet theories, biases, errors
1064 (and good judgment, too) of Google search engineers and product managers
1065 would not have such an outsized effect on consumer choice.
1066   </p><p>
1067     This goes for many other companies. Amazon, a classic surveillance
1068 capitalist, is obviously the dominant tool for searching Amazon — though
1069 many people find their way to Amazon through Google searches and Facebook
1070 posts — and obviously, Amazon controls Amazon search.  That means that
1071 Amazon’s own self-serving editorial choices—like promoting its own house
1072 brands over rival goods from its sellers as well as its own pet theories,
1073 biases, and errors— determine much of what we buy on Amazon. And since
1074 Amazon is the dominant e-commerce retailer outside of China and since it
1075 attained that dominance by buying up both large rivals and nascent
1076 competitors in defiance of historical antitrust rules, we can blame the
1077 monopoly for stripping consumers of their right to the future tense and the
1078 ability to shape markets by making informed choices.
1079   </p><p>
1080     Not every monopolist is a surveillance capitalist, but that doesn’t mean
1081 they’re not able to shape consumer choices in wide-ranging ways. Zuboff
1082 lauds Apple for its App Store and iTunes Store, insisting that adding price
1083 tags to the features on its platforms has been the secret to resisting
1084 surveillance and thus creating markets. But Apple is the only retailer
1085 allowed to sell on its platforms, and it’s the second-largest mobile device
1086 vendor in the world. The independent software vendors that sell through
1087 Apple’s marketplace accuse the company of the same surveillance sins as
1088 Amazon and other big retailers: spying on its customers to find lucrative
1089 new products to launch, effectively using independent software vendors as
1090 free-market researchers, then forcing them out of any markets they discover.
1091   </p><p>
1092     Because of its use of copyright locks, Apple’s mobile customers are not
1093 legally allowed to switch to a rival retailer for its apps if they want to
1094 do so on an iPhone. Apple, obviously, is the only entity that gets to decide
1095 how it ranks the results of search queries in its stores. These decisions
1096 ensure that some apps are often installed (because they appear on page one)
1097 and others are never installed (because they appear on page one
1098 million). Apple’s search-ranking design decisions have a vastly more
1099 significant effect on consumer behaviors than influence campaigns delivered
1100 by surveillance capitalism’s ad-serving bots.
1101   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="monopolists-can-afford-sleeping-pills-for-watchdogs"></a>Monopolists can afford sleeping pills for watchdogs</h2></div></div></div><p>
1102     Only the most extreme market ideologues think that markets can self-regulate
1103 without state oversight. Markets need watchdogs — regulators, lawmakers, and
1104 other elements of democratic control — to keep them honest. When these
1105 watchdogs sleep on the job, then markets cease to aggregate consumer choices
1106 because those choices are constrained by illegitimate and deceptive
1107 activities that companies are able to get away with because no one is
1108 holding them to account.
1109   </p><p>
1110     But this kind of regulatory capture doesn’t come cheap. In competitive
1111 sectors, where rivals are constantly eroding one another’s margins,
1112 individual firms lack the surplus capital to effectively lobby for laws and
1113 regulations that serve their ends.
1114   </p><p>
1115     Many of the harms of surveillance capitalism are the result of weak or
1116 nonexistent regulation. Those regulatory vacuums spring from the power of
1117 monopolists to resist stronger regulation and to tailor what regulation
1118 exists to permit their existing businesses.
1119   </p><p>
1120     Here’s an example: When firms over-collect and over-retain our data, they
1121 are at increased risk of suffering a breach — you can’t leak data you never
1122 collected, and once you delete all copies of that data, you can no longer
1123 leak it. For more than a decade, we’ve lived through an endless parade of
1124 ever-worsening data breaches, each one uniquely horrible in the scale of
1125 data breached and the sensitivity of that data.
1126   </p><p>
1127     But still, firms continue to over-collect and over-retain our data for three
1128 reasons:
1129   </p><p>
1130     <span class="strong"><strong>1. They are locked in the aforementioned limbic arms
1131 race with our capacity to shore up our attentional defense systems to resist
1132 their new persuasion techniques.</strong></span> They’re also locked in an arms
1133 race with their competitors to find new ways to target people for sales
1134 pitches. As soon as they discover a soft spot in our attentional defenses (a
1135 counterintuitive, unobvious way to target potential refrigerator buyers),
1136 the public begins to wise up to the tactic, and their competitors leap on
1137 it, hastening the day in which all potential refrigerator buyers have been
1138 inured to the pitch.
1139   </p><p>
1140     <span class="strong"><strong>2. They believe the surveillance capitalism
1141 story.</strong></span> Data is cheap to aggregate and store, and both proponents
1142 and opponents of surveillance capitalism have assured managers and product
1143 designers that if you collect enough data, you will be able to perform
1144 sorcerous acts of mind control, thus supercharging your sales. Even if you
1145 never figure out how to profit from the data, someone else will eventually
1146 offer to buy it from you to give it a try. This is the hallmark of all
1147 economic bubbles: acquiring an asset on the assumption that someone else
1148 will buy it from you for more than you paid for it, often to sell to someone
1149 else at an even greater price.
1150   </p><p>
1151     <span class="strong"><strong>3. The penalties for leaking data are
1152 negligible.</strong></span> Most countries limit these penalties to actual
1153 damages, meaning that consumers who’ve had their data breached have to show
1154 actual monetary harms to get a reward. In 2014, Home Depot disclosed that it
1155 had lost credit-card data for 53 million of its customers, but it settled
1156 the matter by paying those customers about $0.34 each — and a third of that
1157 $0.34 wasn’t even paid in cash. It took the form of a credit to procure a
1158 largely ineffectual credit-monitoring service.
1159   </p><p>
1160     But the harms from breaches are much more extensive than these
1161 actual-damages rules capture. Identity thieves and fraudsters are wily and
1162 endlessly inventive. All the vast breaches of our century are being
1163 continuously recombined, the data sets merged and mined for new ways to
1164 victimize the people whose data was present in them.  Any reasonable,
1165 evidence-based theory of deterrence and compensation for breaches would not
1166 confine damages to actual damages but rather would allow users to claim
1167 these future harms.
1168   </p><p>
1169     However, even the most ambitious privacy rules, such as the EU General Data
1170 Protection Regulation, fall far short of capturing the negative
1171 externalities of the platforms’ negligent over-collection and
1172 over-retention, and what penalties they do provide are not aggressively
1173 pursued by regulators.
1174   </p><p>
1175     This tolerance of — or indifference to — data over-collection and
1176 over-retention can be ascribed in part to the sheer lobbying muscle of the
1177 platforms. They are so profitable that they can handily afford to divert
1178 gigantic sums to fight any real change — that is, change that would force
1179 them to internalize the costs of their surveillance activities.
1180   </p><p>
1181     And then there’s state surveillance, which the surveillance capitalism story
1182 dismisses as a relic of another era when the big worry was being jailed for
1183 your dissident speech, not having your free will stripped away with machine
1184 learning.
1185   </p><p>
1186     But state surveillance and private surveillance are intimately related. As
1187 we saw when Apple was conscripted by the Chinese government as a vital
1188 collaborator in state surveillance, the only really affordable and tractable
1189 way to conduct mass surveillance on the scale practiced by modern states —
1190 both <span class="quote">« <span class="quote">free</span> »</span> and autocratic states — is to suborn commercial
1191 services.
1192   </p><p>
1193     Whether it’s Google being used as a location tracking tool by local law
1194 enforcement across the U.S. or the use of social media tracking by the
1195 Department of Homeland Security to build dossiers on participants in
1196 protests against Immigration and Customs Enforcement’s family separation
1197 practices, any hard limits on surveillance capitalism would hamstring the
1198 state’s own surveillance capability. Without Palantir, Amazon, Google, and
1199 other major tech contractors, U.S. cops would not be able to spy on Black
1200 people, ICE would not be able to manage the caging of children at the U.S.
1201 border, and state welfare systems would not be able to purge their rolls by
1202 dressing up cruelty as empiricism and claiming that poor and vulnerable
1203 people are ineligible for assistance. At least some of the states’
1204 unwillingness to take meaningful action to curb surveillance should be
1205 attributed to this symbiotic relationship.  There is no mass state
1206 surveillance without mass commercial surveillance.
1207   </p><p>
1208     Monopolism is key to the project of mass state surveillance. It’s true that
1209 smaller tech firms are apt to be less well-defended than Big Tech, whose
1210 security experts are drawn from the tops of their field and who are given
1211 enormous resources to secure and monitor their systems against
1212 intruders. But smaller firms also have less to protect: fewer users whose
1213 data is more fragmented across more systems and have to be suborned one at a
1214 time by state actors.
1215   </p><p>
1216     A concentrated tech sector that works with authorities is a much more
1217 powerful ally in the project of mass state surveillance than a fragmented
1218 one composed of smaller actors. The U.S. tech sector is small enough that
1219 all of its top executives fit around a single boardroom table in Trump Tower
1220 in 2017, shortly after Trump’s inauguration. Most of its biggest players bid
1221 to win JEDI, the Pentagon’s $10 billion Joint Enterprise Defense
1222 Infrastructure cloud contract. Like other highly concentrated industries,
1223 Big Tech rotates its key employees in and out of government service, sending
1224 them to serve in the Department of Defense and the White House, then hiring
1225 ex-Pentagon and ex-DOD top staffers and officers to work in their own
1226 government relations departments.
1227   </p><p>
1228     They can even make a good case for doing this: After all, when there are
1229 only four or five big companies in an industry, everyone qualified to
1230 regulate those companies has served as an executive in at least a couple of
1231 them — because, likewise, when there are only five companies in an industry,
1232 everyone qualified for a senior role at any of them is by definition working
1233 at one of the other ones.
1234   </p><div class="blockquote"><blockquote class="blockquote"><p>
1235       While surveillance doesn’t cause monopolies, monopolies certainly abet
1236 surveillance.
1237     </p></blockquote></div><p>
1238     Industries that are competitive are fragmented — composed of companies that
1239 are at each other’s throats all the time and eroding one another’s margins
1240 in bids to steal their best customers. This leaves them with much more
1241 limited capital to use to lobby for favorable rules and a much harder job of
1242 getting everyone to agree to pool their resources to benefit the industry as
1243 a whole.
1244   </p><p>
1245     Surveillance combined with machine learning is supposed to be an existential
1246 crisis, a species-defining moment at which our free will is just a few more
1247 advances in the field from being stripped away. I am skeptical of this
1248 claim, but I <span class="emphasis"><em>do</em></span> think that tech poses an existential
1249 threat to our society and possibly our species.
1250   </p><p>
1251     But that threat grows out of monopoly.
1252   </p><p>
1253     One of the consequences of tech’s regulatory capture is that it can shift
1254 liability for poor security decisions onto its customers and the wider
1255 society. It is absolutely normal in tech for companies to obfuscate the
1256 workings of their products, to make them deliberately hard to understand,
1257 and to threaten security researchers who seek to independently audit those
1258 products.
1259   </p><p>
1260     IT is the only field in which this is practiced: No one builds a bridge or a
1261 hospital and keeps the composition of the steel or the equations used to
1262 calculate load stresses a secret. It is a frankly bizarre practice that
1263 leads, time and again, to grotesque security defects on farcical scales,
1264 with whole classes of devices being revealed as vulnerable long after they
1265 are deployed in the field and put into sensitive places.
1266   </p><p>
1267     The monopoly power that keeps any meaningful consequences for breaches at
1268 bay means that tech companies continue to build terrible products that are
1269 insecure by design and that end up integrated into our lives, in possession
1270 of our data, and connected to our physical world. For years, Boeing has
1271 struggled with the aftermath of a series of bad technology decisions that
1272 made its 737 fleet a global pariah, a rare instance in which bad tech
1273 decisions have been seriously punished in the market.
1274   </p><p>
1275     These bad security decisions are compounded yet again by the use of
1276 copyright locks to enforce business-model decisions against
1277 consumers. Recall that these locks have become the go-to means for shaping
1278 consumer behavior, making it technically impossible to use third-party ink,
1279 insulin, apps, or service depots in connection with your lawfully acquired
1280 property.
1281   </p><p>
1282     Recall also that these copyright locks are backstopped by legislation (such
1283 as Section 1201 of the DMCA or Article 6 of the 2001 EU Copyright Directive)
1284 that ban tampering with (<span class="quote">« <span class="quote">circumventing</span> »</span>) them, and these
1285 statutes have been used to threaten security researchers who make
1286 disclosures about vulnerabilities without permission from manufacturers.
1287   </p><p>
1288     This amounts to a manufacturer’s veto over safety warnings and
1289 criticism. While this is far from the legislative intent of the DMCA and its
1290 sister statutes around the world, Congress has not intervened to clarify the
1291 statute nor will it because to do so would run counter to the interests of
1292 powerful, large firms whose lobbying muscle is unstoppable.
1293   </p><p>
1294     Copyright locks are a double whammy: They create bad security decisions that
1295 can’t be freely investigated or discussed. If markets are supposed to be
1296 machines for aggregating information (and if surveillance capitalism’s
1297 notional mind-control rays are what make it a <span class="quote">« <span class="quote">rogue
1298 capitalism</span> »</span> because it denies consumers the power to make decisions),
1299 then a program of legally enforced ignorance of the risks of products makes
1300 monopolism even more of a <span class="quote">« <span class="quote">rogue capitalism</span> »</span> than surveillance
1301 capitalism’s influence campaigns.
1302   </p><p>
1303     And unlike mind-control rays, enforced silence over security is an
1304 immediate, documented problem, and it <span class="emphasis"><em>does</em></span> constitute
1305 an existential threat to our civilization and possibly our species. The
1306 proliferation of insecure devices — especially devices that spy on us and
1307 especially when those devices also can manipulate the physical world by,
1308 say, steering your car or flipping a breaker at a power station — is a kind
1309 of technology debt.
1310   </p><p>
1311     In software design, <span class="quote">« <span class="quote">technology debt</span> »</span> refers to old, baked-in
1312 decisions that turn out to be bad ones in hindsight. Perhaps a long-ago
1313 developer decided to incorporate a networking protocol made by a vendor that
1314 has since stopped supporting it. But everything in the product still relies
1315 on that superannuated protocol, and so, with each revision, the product team
1316 has to work around this obsolete core, adding compatibility layers,
1317 surrounding it with security checks that try to shore up its defenses, and
1318 so on. These Band-Aid measures compound the debt because every subsequent
1319 revision has to make allowances for <span class="emphasis"><em>them</em></span>, too, like
1320 interest mounting on a predatory subprime loan. And like a subprime loan,
1321 the interest mounts faster than you can hope to pay it off: The product team
1322 has to put so much energy into maintaining this complex, brittle system that
1323 they don’t have any time left over to refactor the product from the ground
1324 up and <span class="quote">« <span class="quote">pay off the debt</span> »</span> once and for all.
1325   </p><p>
1326     Typically, technology debt results in a technological bankruptcy: The
1327 product gets so brittle and unsustainable that it fails
1328 catastrophically. Think of the antiquated COBOL-based banking and accounting
1329 systems that fell over at the start of the pandemic emergency when
1330 confronted with surges of unemployment claims.  Sometimes that ends the
1331 product; sometimes it takes the company down with it. Being caught in the
1332 default of a technology debt is scary and traumatic, just like losing your
1333 house due to bankruptcy is scary and traumatic.
1334   </p><p>
1335     But the technology debt created by copyright locks isn’t individual debt;
1336 it’s systemic. Everyone in the world is exposed to this over-leverage, as
1337 was the case with the 2008 financial crisis. When that debt comes due — when
1338 we face a cascade of security breaches that threaten global shipping and
1339 logistics, the food supply, pharmaceutical production pipelines, emergency
1340 communications, and other critical systems that are accumulating technology
1341 debt in part due to the presence of deliberately insecure and deliberately
1342 unauditable copyright locks — it will indeed pose an existential risk.
1343   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="privacy-and-monopoly"></a>Privacy and monopoly</h2></div></div></div><p>
1344     Many tech companies are gripped by an orthodoxy that holds that if they just
1345 gather enough data on enough of our activities, everything else is possible
1346 — the mind control and endless profits. This is an unfalsifiable hypothesis:
1347 If data gives a tech company even a tiny improvement in behavior prediction
1348 and modification, the company declares that it has taken the first step
1349 toward global domination with no end in sight. If a company
1350 <span class="emphasis"><em>fails</em></span> to attain any improvements from gathering and
1351 analyzing data, it declares success to be just around the corner, attainable
1352 once more data is in hand.
1353   </p><p>
1354     Surveillance tech is far from the first industry to embrace a nonsensical,
1355 self-serving belief that harms the rest of the world, and it is not the
1356 first industry to profit handsomely from such a delusion. Long before
1357 hedge-fund managers were claiming (falsely)  that they could beat the
1358 S&amp;P 500, there were plenty of other <span class="quote">« <span class="quote">respectable</span> »</span>
1359 industries that have been revealed as quacks in hindsight. From the makers
1360 of radium suppositories (a real thing!)  to the cruel sociopaths who claimed
1361 they could <span class="quote">« <span class="quote">cure</span> »</span> gay people, history is littered with the
1362 formerly respectable titans of discredited industries.
1363   </p><p>
1364     This is not to say that there’s nothing wrong with Big Tech and its
1365 ideological addiction to data. While surveillance’s benefits are mostly
1366 overstated, its harms are, if anything, <span class="emphasis"><em>understated</em></span>.
1367   </p><p>
1368     There’s real irony here. The belief in surveillance capitalism as a
1369 <span class="quote">« <span class="quote">rogue capitalism</span> »</span> is driven by the belief that markets
1370 wouldn’t tolerate firms that are gripped by false beliefs. An oil company
1371 that has false beliefs about where the oil is will eventually go broke
1372 digging dry wells after all.
1373   </p><p>
1374     But monopolists get to do terrible things for a long time before they pay
1375 the price. Think of how concentration in the finance sector allowed the
1376 subprime crisis to fester as bond-rating agencies, regulators, investors,
1377 and critics all fell under the sway of a false belief that complex
1378 mathematics could construct <span class="quote">« <span class="quote">fully hedged</span> »</span> debt instruments
1379 that could not possibly default. A small bank that engaged in this kind of
1380 malfeasance would simply go broke rather than outrunning the inevitable
1381 crisis, perhaps growing so big that it averted it altogether. But large
1382 banks were able to continue to attract investors, and when they finally
1383 <span class="emphasis"><em>did</em></span> come a-cropper, the world’s governments bailed them
1384 out. The worst offenders of the subprime crisis are bigger than they were in
1385 2008, bringing home more profits and paying their execs even larger sums.
1386   </p><p>
1387     Big Tech is able to practice surveillance not just because it is tech but
1388 because it is <span class="emphasis"><em>big</em></span>. The reason every web publisher
1389 embeds a Facebook <span class="quote">« <span class="quote">Like</span> »</span> button is that Facebook dominates the
1390 internet’s social media referrals — and every one of those
1391 <span class="quote">« <span class="quote">Like</span> »</span> buttons spies on everyone who lands on a page that
1392 contains them (see also: Google Analytics embeds, Twitter buttons, etc.).
1393   </p><p>
1394     The reason the world’s governments have been slow to create meaningful
1395 penalties for privacy breaches is that Big Tech’s concentration produces
1396 huge profits that can be used to lobby against those penalties — and Big
1397 Tech’s concentration means that the companies involved are able to arrive at
1398 a unified negotiating position that supercharges the lobbying.
1399   </p><p>
1400     The reason that the smartest engineers in the world want to work for Big
1401 Tech is that Big Tech commands the lion’s share of tech industry jobs.
1402   </p><p>
1403     The reason people who are aghast at Facebook’s and Google’s and Amazon’s
1404 data-handling practices continue to use these services is that all their
1405 friends are on Facebook; Google dominates search; and Amazon has put all the
1406 local merchants out of business.
1407   </p><p>
1408     Competitive markets would weaken the companies’ lobbying muscle by reducing
1409 their profits and pitting them against each other in regulatory forums. It
1410 would give customers other places to go to get their online services. It
1411 would make the companies small enough to regulate and pave the way to
1412 meaningful penalties for breaches. It would let engineers with ideas that
1413 challenged the surveillance orthodoxy raise capital to compete with the
1414 incumbents. It would give web publishers multiple ways to reach audiences
1415 and make the case against Facebook and Google and Twitter embeds.
1416   </p><p>
1417     In other words, while surveillance doesn’t cause monopolies, monopolies
1418 certainly abet surveillance.
1419   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="ronald-reagan-pioneer-of-tech-monopolism"></a>Ronald Reagan, pioneer of tech monopolism</h2></div></div></div><p>
1420     Technology exceptionalism is a sin, whether it’s practiced by technology’s
1421 blind proponents or by its critics. Both of these camps are prone to
1422 explaining away monopolistic concentration by citing some special
1423 characteristic of the tech industry, like network effects or first-mover
1424 advantage. The only real difference between these two groups is that the
1425 tech apologists say monopoly is inevitable so we should just let tech get
1426 away with its abuses while competition regulators in the U.S. and the EU say
1427 monopoly is inevitable so we should punish tech for its abuses but not try
1428 to break up the monopolies.
1429   </p><p>
1430     To understand how tech became so monopolistic, it’s useful to look at the
1431 dawn of the consumer tech industry: 1979, the year the Apple II Plus
1432 launched and became the first successful home computer. That also happens to
1433 be the year that Ronald Reagan hit the campaign trail for the 1980
1434 presidential race — a race he won, leading to a radical shift in the way
1435 that antitrust concerns are handled in America. Reagan’s cohort of
1436 politicians — including Margaret Thatcher in the U.K., Brian Mulroney in
1437 Canada, Helmut Kohl in Germany, and Augusto Pinochet in Chile — went on to
1438 enact similar reforms that eventually spread around the world.
1439   </p><p>
1440     Antitrust’s story began nearly a century before all that with laws like the
1441 Sherman Act, which took aim at monopolists on the grounds that monopolies
1442 were bad in and of themselves — squeezing out competitors, creating
1443 <span class="quote">« <span class="quote">diseconomies of scale</span> »</span> (when a company is so big that its
1444 constituent parts go awry and it is seemingly helpless to address the
1445 problems), and capturing their regulators to such a degree that they can get
1446 away with a host of evils.
1447   </p><p>
1448     Then came a fabulist named Robert Bork, a former solicitor general who
1449 Reagan appointed to the powerful U.S. Court of Appeals for the D.C. Circuit
1450 and who had created an alternate legislative history of the Sherman Act and
1451 its successors out of whole cloth. Bork insisted that these statutes were
1452 never targeted at monopolies (despite a wealth of evidence to the contrary,
1453 including the transcribed speeches of the acts’ authors) but, rather, that
1454 they were intended to prevent <span class="quote">« <span class="quote">consumer harm</span> »</span> — in the form of
1455 higher prices.
1456   </p><p>
1457     Bork was a crank, but he was a crank with a theory that rich people really
1458 liked. Monopolies are a great way to make rich people richer by allowing
1459 them to receive <span class="quote">« <span class="quote">monopoly rents</span> »</span> (that is, bigger profits) and
1460 capture regulators, leading to a weaker, more favorable regulatory
1461 environment with fewer protections for customers, suppliers, the
1462 environment, and workers.
1463   </p><p>
1464     Bork’s theories were especially palatable to the same power brokers who
1465 backed Reagan, and Reagan’s Department of Justice and other agencies began
1466 to incorporate Bork’s antitrust doctrine into their enforcement decisions
1467 (Reagan even put Bork up for a Supreme Court seat, but Bork flunked the
1468 Senate confirmation hearing so badly that, 40 years later, D.C. insiders use
1469 the term <span class="quote">« <span class="quote">borked</span> »</span> to refer to any catastrophically bad
1470 political performance).
1471   </p><p>
1472     Little by little, Bork’s theories entered the mainstream, and their backers
1473 began to infiltrate the legal education field, even putting on junkets where
1474 members of the judiciary were treated to lavish meals, fun outdoor
1475 activities, and seminars where they were indoctrinated into the consumer
1476 harm theory of antitrust. The more Bork’s theories took hold, the more money
1477 the monopolists were making — and the more surplus capital they had at their
1478 disposal to lobby for even more Borkian antitrust influence campaigns.
1479   </p><p>
1480     The history of Bork’s antitrust theories is a really good example of the
1481 kind of covertly engineered shifts in public opinion that Zuboff warns us
1482 against, where fringe ideas become mainstream orthodoxy.  But Bork didn’t
1483 change the world overnight. He played a very long game, for over a
1484 generation, and he had a tailwind because the same forces that backed
1485 oligarchic antitrust theories also backed many other oligarchic shifts in
1486 public opinion. For example, the idea that taxation is theft, that wealth is
1487 a sign of virtue, and so on — all of these theories meshed to form a
1488 coherent ideology that elevated inequality to a virtue.
1489   </p><p>
1490     Today, many fear that machine learning allows surveillance capitalism to
1491 sell <span class="quote">« <span class="quote">Bork-as-a-Service,</span> »</span> at internet speeds, so that you can
1492 contract a machine-learning company to engineer <span class="emphasis"><em>rapid</em></span>
1493 shifts in public sentiment without needing the capital to sustain a
1494 multipronged, multigenerational project working at the local, state,
1495 national, and global levels in business, law, and philosophy. I do not
1496 believe that such a project is plausible, though I agree that this is
1497 basically what the platforms claim to be selling. They’re just lying about
1498 it. Big Tech lies all the time, <span class="emphasis"><em>including</em></span> in their
1499 sales literature.
1500   </p><p>
1501     The idea that tech forms <span class="quote">« <span class="quote">natural monopolies</span> »</span> (monopolies that
1502 are the inevitable result of the realities of an industry, such as the
1503 monopolies that accrue the first company to run long-haul phone lines or
1504 rail lines) is belied by tech’s own history: In the absence of
1505 anti-competitive tactics, Google was able to unseat AltaVista and Yahoo;
1506 Facebook was able to head off Myspace. There are some advantages to
1507 gathering mountains of data, but those mountains of data also have
1508 disadvantages: liability (from leaking), diminishing returns (from old
1509 data), and institutional inertia (big companies, like science, progress one
1510 funeral at a time).
1511   </p><p>
1512     Indeed, the birth of the web saw a mass-extinction event for the existing
1513 giant, wildly profitable proprietary technologies that had capital, network
1514 effects, and walls and moats surrounding their businesses. The web showed
1515 that when a new industry is built around a protocol, rather than a product,
1516 the combined might of everyone who uses the protocol to reach their
1517 customers or users or communities outweighs even the most massive
1518 products. CompuServe, AOL, MSN, and a host of other proprietary walled
1519 gardens learned this lesson the hard way: Each believed it could stay
1520 separate from the web, offering <span class="quote">« <span class="quote">curation</span> »</span> and a guarantee of
1521 consistency and quality instead of the chaos of an open system. Each was
1522 wrong and ended up being absorbed into the public web.
1523   </p><p>
1524     Yes, tech is heavily monopolized and is now closely associated with industry
1525 concentration, but this has more to do with a matter of timing than its
1526 intrinsically monopolistic tendencies. Tech was born at the moment that
1527 antitrust enforcement was being dismantled, and tech fell into exactly the
1528 same pathologies that antitrust was supposed to guard against. To a first
1529 approximation, it is reasonable to assume that tech’s monopolies are the
1530 result of a lack of anti-monopoly action and not the much-touted unique
1531 characteristics of tech, such as network effects, first-mover advantage, and
1532 so on.
1533   </p><p>
1534     In support of this thesis, I offer the concentration that every
1535 <span class="emphasis"><em>other</em></span> industry has undergone over the same period. From
1536 professional wrestling to consumer packaged goods to commercial property
1537 leasing to banking to sea freight to oil to record labels to newspaper
1538 ownership to theme parks, <span class="emphasis"><em>every</em></span> industry has undergone
1539 a massive shift toward concentration. There’s no obvious network effects or
1540 first-mover advantage at play in these industries. However, in every case,
1541 these industries attained their concentrated status through tactics that
1542 were prohibited before Bork’s triumph: merging with major competitors,
1543 buying out innovative new market entrants, horizontal and vertical
1544 integration, and a suite of anti-competitive tactics that were once illegal
1545 but are not any longer.
1546   </p><p>
1547     Again: When you change the laws intended to prevent monopolies and then
1548 monopolies form in exactly the way the law was supposed to prevent, it is
1549 reasonable to suppose that these facts are related.  Tech’s concentration
1550 can be readily explained without recourse to radical theories of network
1551 effects — but only if you’re willing to indict unregulated markets as
1552 tending toward monopoly. Just as a lifelong smoker can give you a hundred
1553 reasons why their smoking didn’t cause their cancer (<span class="quote">« <span class="quote">It was the
1554 environmental toxins</span> »</span>), true believers in unregulated markets have a
1555 whole suite of unconvincing explanations for monopoly in tech that leave
1556 capitalism intact.
1557   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="steering-with-the-windshield-wipers"></a>Steering with the windshield wipers</h2></div></div></div><p>
1558     It’s been 40 years since Bork’s project to rehabilitate monopolies achieved
1559 liftoff, and that is a generation and a half, which is plenty of time to
1560 take a common idea and make it seem outlandish and vice versa. Before the
1561 1940s, affluent Americans dressed their baby boys in pink while baby girls
1562 wore blue (a <span class="quote">« <span class="quote">delicate and dainty</span> »</span> color). While gendered
1563 colors are obviously totally arbitrary, many still greet this news with
1564 amazement and find it hard to imagine a time when pink connoted masculinity.
1565   </p><p>
1566     After 40 years of studiously ignoring antitrust analysis and enforcement,
1567 it’s not surprising that we’ve all but forgotten that antitrust exists, that
1568 in living memory, growth through mergers and acquisitions were largely
1569 prohibited under law, that market-cornering strategies like vertical
1570 integration could land a company in court.
1571   </p><p>
1572     Antitrust is a market society’s steering wheel, the control of first resort
1573 to keep would-be masters of the universe in their lanes. But Bork and his
1574 cohort ripped out our steering wheel 40 years ago. The car is still
1575 barreling along, and so we’re yanking as hard as we can on all the
1576 <span class="emphasis"><em>other</em></span> controls in the car as well as desperately
1577 flapping the doors and rolling the windows up and down in the hopes that one
1578 of these other controls can be repurposed to let us choose where we’re
1579 heading before we careen off a cliff.
1580   </p><p>
1581     It’s like a 1960s science-fiction plot come to life: People stuck in a
1582 <span class="quote">« <span class="quote">generation ship,</span> »</span> plying its way across the stars, a ship once
1583 piloted by their ancestors; and now, after a great cataclysm, the ship’s
1584 crew have forgotten that they’re in a ship at all and no longer remember
1585 where the control room is. Adrift, the ship is racing toward its extinction,
1586 and unless we can seize the controls and execute emergency course
1587 correction, we’re all headed for a fiery death in the heart of a sun.
1588   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="surveillance-still-matters"></a>Surveillance still matters</h2></div></div></div><p>
1589     None of this is to minimize the problems with surveillance.  Surveillance
1590 matters, and Big Tech’s use of surveillance <span class="emphasis"><em>is</em></span> an
1591 existential risk to our species, but that’s not because surveillance and
1592 machine learning rob us of our free will.
1593   </p><p>
1594     Surveillance has become <span class="emphasis"><em>much</em></span> more efficient thanks to
1595 Big Tech. In 1989, the Stasi — the East German secret police — had the whole
1596 country under surveillance, a massive undertaking that recruited one out of
1597 every 60 people to serve as an informant or intelligence operative.
1598   </p><p>
1599     Today, we know that the NSA is spying on a significant fraction of the
1600 entire world’s population, and its ratio of surveillance operatives to the
1601 surveilled is more like 1:10,000 (that’s probably on the low side since it
1602 assumes that every American with top-secret clearance is working for the NSA
1603 on this project — we don’t know how many of those cleared people are
1604 involved in NSA spying, but it’s definitely not all of them).
1605   </p><p>
1606     How did the ratio of surveillable citizens expand from 1:60 to 1:10,000 in
1607 less than 30 years? It’s thanks to Big Tech. Our devices and services gather
1608 most of the data that the NSA mines for its surveillance project. We pay for
1609 these devices and the services they connect to, and then we painstakingly
1610 perform the data-entry tasks associated with logging facts about our lives,
1611 opinions, and preferences. This mass surveillance project has been largely
1612 useless for fighting terrorism: The NSA can <a class="ulink" href="https://www.washingtonpost.com/world/national-security/nsa-cites-case-as-success-of-phone-data-collection-program/2013/08/08/fc915e5a-feda-11e2-96a8-d3b921c0924a_story.html" target="_top">only
1613 point to a single minor success story</a> in which it used its data
1614 collection program to foil an attempt by a U.S. resident to wire a few
1615 thousand dollars to an overseas terror group. It’s ineffective for much the
1616 same reason that commercial surveillance projects are largely ineffective at
1617 targeting advertising: The people who want to commit acts of terror, like
1618 people who want to buy a refrigerator, are extremely rare. If you’re trying
1619 to detect a phenomenon whose base rate is one in a million with an
1620 instrument whose accuracy is only 99%, then every true positive will come at
1621 the cost of 9,999 false positives.
1622   </p><p>
1623     Let me explain that again: If one in a million people is a terrorist, then
1624 there will only be about one terrorist in a random sample of one million
1625 people. If your test for detecting terrorists is 99% accurate, it will
1626 identify 10,000 terrorists in your million-person sample (1% of one million
1627 is 10,000). For every true positive, you’ll get 9,999 false positives.
1628   </p><p>
1629     In reality, the accuracy of algorithmic terrorism detection falls far short
1630 of the 99% mark, as does refrigerator ad targeting. The difference is that
1631 being falsely accused of wanting to buy a fridge is a minor nuisance while
1632 being falsely accused of planning a terror attack can destroy your life and
1633 the lives of everyone you love.
1634   </p><p>
1635     Mass state surveillance is only feasible because of surveillance capitalism
1636 and its extremely low-yield ad-targeting systems, which require a constant
1637 feed of personal data to remain barely viable.  Surveillance capitalism’s
1638 primary failure mode is mistargeted ads while mass state surveillance’s
1639 primary failure mode is grotesque human rights abuses, tending toward
1640 totalitarianism.
1641   </p><p>
1642     State surveillance is no mere parasite on Big Tech, sucking up its data and
1643 giving nothing in return. In truth, the two are symbiotes: Big Tech sucks up
1644 our data for spy agencies, and spy agencies ensure that governments don’t
1645 limit Big Tech’s activities so severely that it would no longer serve the
1646 spy agencies’ needs. There is no firm distinction between state surveillance
1647 and surveillance capitalism; they are dependent on one another.
1648   </p><p>
1649     To see this at work today, look no further than Amazon’s home surveillance
1650 device, the Ring doorbell, and its associated app, Neighbors. Ring — a
1651 product that Amazon acquired and did not develop in house — makes a
1652 camera-enabled doorbell that streams footage from your front door to your
1653 mobile device. The Neighbors app allows you to form a neighborhood-wide
1654 surveillance grid with your fellow Ring owners through which you can share
1655 clips of <span class="quote">« <span class="quote">suspicious characters.</span> »</span> If you’re thinking that this
1656 sounds like a recipe for letting curtain-twitching racists supercharge their
1657 suspicions of people with brown skin who walk down their blocks, <a class="ulink" href="https://www.eff.org/deeplinks/2020/07/amazons-ring-enables-over-policing-efforts-some-americas-deadliest-law-enforcement" target="_top">you’re
1658 right</a>. Ring has become a <span class="emphasis"><em>de facto,</em></span>
1659 off-the-books arm of the police without any of the pesky oversight or rules.
1660   </p><p>
1661     In mid-2019, a series of public records requests revealed that Amazon had
1662 struck confidential deals with more than 400 local law enforcement agencies
1663 through which the agencies would promote Ring and Neighbors and in exchange
1664 get access to footage from Ring cameras. In theory, cops would need to
1665 request this footage through Amazon (and internal documents reveal that
1666 Amazon devotes substantial resources to coaching cops on how to spin a
1667 convincing story when doing so), but in practice, when a Ring customer turns
1668 down a police request, Amazon only requires the agency to formally request
1669 the footage from the company, which it will then produce.
1670   </p><p>
1671     Ring and law enforcement have found many ways to intertwine their
1672 activities. Ring strikes secret deals to acquire real-time access to 911
1673 dispatch and then streams alarming crime reports to Neighbors users, which
1674 serve as convincers for anyone who’s contemplating a surveillance doorbell
1675 but isn’t sure whether their neighborhood is dangerous enough to warrant it.
1676   </p><p>
1677     The more the cops buzz-market the surveillance capitalist Ring, the more
1678 surveillance capability the state gets. Cops who rely on private entities
1679 for law-enforcement roles then brief against any controls on the deployment
1680 of that technology while the companies return the favor by lobbying against
1681 rules requiring public oversight of police surveillance technology. The more
1682 the cops rely on Ring and Neighbors, the harder it will be to pass laws to
1683 curb them. The fewer laws there are against them, the more the cops will
1684 rely on them.
1685   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="dignity-and-sanctuary"></a>Dignity and sanctuary</h2></div></div></div><p>
1686     But even if we could exercise democratic control over our states and force
1687 them to stop raiding surveillance capitalism’s reservoirs of behavioral
1688 data, surveillance capitalism would still harm us.
1689   </p><p>
1690     This is an area where Zuboff shines. Her chapter on <span class="quote">« <span class="quote">sanctuary</span> »</span>
1691 — the feeling of being unobserved — is a beautiful hymn to introspection,
1692 calmness, mindfulness, and tranquility.
1693   </p><p>
1694     When you are watched, something changes. Anyone who has ever raised a child
1695 knows this. You might look up from your book (or more realistically, from
1696 your phone) and catch your child in a moment of profound realization and
1697 growth, a moment where they are learning something that is right at the edge
1698 of their abilities, requiring their entire ferocious concentration. For a
1699 moment, you’re transfixed, watching that rare and beautiful moment of focus
1700 playing out before your eyes, and then your child looks up and sees you
1701 seeing them, and the moment collapses. To grow, you need to be and expose
1702 your authentic self, and in that moment, you are vulnerable like a hermit
1703 crab scuttling from one shell to the next. The tender, unprotected tissues
1704 you expose in that moment are too delicate to reveal in the presence of
1705 another, even someone you trust as implicitly as a child trusts their
1706 parent.
1707   </p><p>
1708     In the digital age, our authentic selves are inextricably tied to our
1709 digital lives. Your search history is a running ledger of the questions
1710 you’ve pondered. Your location history is a record of the places you’ve
1711 sought out and the experiences you’ve had there. Your social graph reveals
1712 the different facets of your identity, the people you’ve connected with.
1713   </p><p>
1714     To be observed in these activities is to lose the sanctuary of your
1715 authentic self.
1716   </p><p>
1717     There’s another way in which surveillance capitalism robs us of our capacity
1718 to be our authentic selves: by making us anxious.  Surveillance capitalism
1719 isn’t really a mind-control ray, but you don’t need a mind-control ray to
1720 make someone anxious. After all, another word for anxiety is agitation, and
1721 to make someone experience agitation, you need merely to agitate them. To
1722 poke them and prod them and beep at them and buzz at them and bombard them
1723 on an intermittent schedule that is just random enough that our limbic
1724 systems never quite become inured to it.
1725   </p><p>
1726     Our devices and services are <span class="quote">« <span class="quote">general purpose</span> »</span> in that they can
1727 connect anything or anyone to anything or anyone else and that they can run
1728 any program that can be written. This means that the distraction rectangles
1729 in our pockets hold our most precious moments with our most beloved people
1730 and their most urgent or time-sensitive communications (from <span class="quote">« <span class="quote">running
1731 late can you get the kid?</span> »</span> to <span class="quote">« <span class="quote">doctor gave me bad news and I
1732 need to talk to you RIGHT NOW</span> »</span>) as well as ads for refrigerators and
1733 recruiting messages from Nazis.
1734   </p><p>
1735     All day and all night, our pockets buzz, shattering our concentration and
1736 tearing apart the fragile webs of connection we spin as we think through
1737 difficult ideas. If you locked someone in a cell and agitated them like
1738 this, we’d call it <span class="quote">« <span class="quote">sleep deprivation torture,</span> »</span> and it would be
1739 <a class="ulink" href="https://www.youtube.com/watch?v=1SKpRbvnx6g" target="_top">a war crime under
1740 the Geneva Conventions</a>.
1741   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="afflicting-the-afflicted"></a>Afflicting the afflicted</h2></div></div></div><p>
1742     The effects of surveillance on our ability to be our authentic selves are
1743 not equal for all people. Some of us are lucky enough to live in a time and
1744 place in which all the most important facts of our lives are widely and
1745 roundly socially acceptable and can be publicly displayed without the risk
1746 of social consequence.
1747   </p><p>
1748     But for many of us, this is not true. Recall that in living memory, many of
1749 the ways of being that we think of as socially acceptable today were once
1750 cause for dire social sanction or even imprisonment.  If you are 65 years
1751 old, you have lived through a time in which people living in <span class="quote">« <span class="quote">free
1752 societies</span> »</span> could be imprisoned or sanctioned for engaging in
1753 homosexual activity, for falling in love with a person whose skin was a
1754 different color than their own, or for smoking weed.
1755   </p><p>
1756     Today, these activities aren’t just decriminalized in much of the world,
1757 they’re considered normal, and the fallen prohibitions are viewed as
1758 shameful, regrettable relics of the past.
1759   </p><p>
1760     How did we get from prohibition to normalization? Through private, personal
1761 activity: People who were secretly gay or secret pot-smokers or who secretly
1762 loved someone with a different skin color were vulnerable to retaliation if
1763 they made their true selves known and were limited in how much they could
1764 advocate for their own right to exist in the world and be true to
1765 themselves. But because there was a private sphere, these people could form
1766 alliances with their friends and loved ones who did not share their
1767 disfavored traits by having private conversations in which they came out,
1768 disclosing their true selves to the people around them and bringing them to
1769 their cause one conversation at a time.
1770   </p><p>
1771     The right to choose the time and manner of these conversations was key to
1772 their success. It’s one thing to come out to your dad while you’re on a
1773 fishing trip away from the world and another thing entirely to blurt it out
1774 over the Christmas dinner table while your racist Facebook uncle is there to
1775 make a scene.
1776   </p><p>
1777     Without a private sphere, there’s a chance that none of these changes would
1778 have come to pass and that the people who benefited from these changes would
1779 have either faced social sanction for coming out to a hostile world or would
1780 have never been able to reveal their true selves to the people they love.
1781   </p><p>
1782     The corollary is that, unless you think that our society has attained social
1783 perfection — that your grandchildren in 50 years will ask you to tell them
1784 the story of how, in 2020, every injustice had been righted and no further
1785 change had to be made — then you should expect that right now, at this
1786 minute, there are people you love, whose happiness is key to your own, who
1787 have a secret in their hearts that stops them from ever being their
1788 authentic selves with you. These people are sorrowing and will go to their
1789 graves with that secret sorrow in their hearts, and the source of that
1790 sorrow will be the falsity of their relationship to you.
1791   </p><p>
1792     A private realm is necessary for human progress.
1793   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="any-data-you-collect-and-retain-will-eventually-leak"></a>Any data you collect and retain will eventually leak</h2></div></div></div><p>
1794     The lack of a private life can rob vulnerable people of the chance to be
1795 their authentic selves and constrain our actions by depriving us of
1796 sanctuary, but there is another risk that is borne by everyone, not just
1797 people with a secret: crime.
1798   </p><p>
1799     Personally identifying information is of very limited use for the purpose of
1800 controlling peoples’ minds, but identity theft — really a catchall term for
1801 a whole constellation of terrible criminal activities that can destroy your
1802 finances, compromise your personal integrity, ruin your reputation, or even
1803 expose you to physical danger — thrives on it.
1804   </p><p>
1805     Attackers are not limited to using data from one breached source,
1806 either. Multiple services have suffered breaches that exposed names,
1807 addresses, phone numbers, passwords, sexual tastes, school grades, work
1808 performance, brushes with the criminal justice system, family details,
1809 genetic information, fingerprints and other biometrics, reading habits,
1810 search histories, literary tastes, pseudonymous identities, and other
1811 sensitive information. Attackers can merge data from these different
1812 breaches to build up extremely detailed dossiers on random subjects and then
1813 use different parts of the data for different criminal purposes.
1814   </p><p>
1815     For example, attackers can use leaked username and password combinations to
1816 hijack whole fleets of commercial vehicles that <a class="ulink" href="https://www.vice.com/en_us/article/zmpx4x/hacker-monitor-cars-kill-engine-gps-tracking-apps" target="_top">have
1817 been fitted with anti-theft GPS trackers and immobilizers</a> or to
1818 hijack baby monitors in order to <a class="ulink" href="https://www.washingtonpost.com/technology/2019/04/23/how-nest-designed-keep-intruders-out-peoples-homes-effectively-allowed-hackers-get/?utm_term=.15220e98c550" target="_top">terrorize
1819 toddlers with the audio tracks from pornography</a>. Attackers use
1820 leaked data to trick phone companies into giving them your phone number,
1821 then they intercept SMS-based two-factor authentication codes in order to
1822 take over your email, bank account, and/or cryptocurrency wallets.
1823   </p><p>
1824     Attackers are endlessly inventive in the pursuit of creative ways to
1825 weaponize leaked data. One common use of leaked data is to penetrate
1826 companies in order to access <span class="emphasis"><em>more</em></span> data.
1827   </p><p>
1828     Like spies, online fraudsters are totally dependent on companies
1829 over-collecting and over-retaining our data. Spy agencies sometimes pay
1830 companies for access to their data or intimidate them into giving it up, but
1831 sometimes they work just like criminals do — by <a class="ulink" href="https://www.bbc.com/news/world-us-canada-24751821" target="_top">sneaking data out of
1832 companies’ databases</a>.
1833   </p><p>
1834     The over-collection of data has a host of terrible social consequences, from
1835 the erosion of our authentic selves to the undermining of social progress,
1836 from state surveillance to an epidemic of online crime. Commercial
1837 surveillance is also a boon to people running influence campaigns, but
1838 that’s the least of our troubles.
1839   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="critical-tech-exceptionalism-is-still-tech-exceptionalism"></a>Critical tech exceptionalism is still tech exceptionalism</h2></div></div></div><p>
1840     Big Tech has long practiced technology exceptionalism: the idea that it
1841 should not be subject to the mundane laws and norms of
1842 <span class="quote">« <span class="quote">meatspace.</span> »</span> Mottoes like Facebook’s <span class="quote">« <span class="quote">move fast and break
1843 things</span> »</span> attracted justifiable scorn of the companies’ self-serving
1844 rhetoric.
1845   </p><p>
1846     Tech exceptionalism got us all into a lot of trouble, so it’s ironic and
1847 distressing to see Big Tech’s critics committing the same sin.
1848   </p><p>
1849     Big Tech is not a <span class="quote">« <span class="quote">rogue capitalism</span> »</span> that cannot be cured
1850 through the traditional anti-monopoly remedies of trustbusting (forcing
1851 companies to divest of competitors they have acquired) and bans on mergers
1852 to monopoly and other anti-competitive tactics. Big Tech does not have the
1853 power to use machine learning to influence our behavior so thoroughly that
1854 markets lose the ability to punish bad actors and reward superior
1855 competitors. Big Tech has no rule-writing mind-control ray that necessitates
1856 ditching our old toolbox.
1857   </p><p>
1858     The thing is, people have been claiming to have perfected mind-control rays
1859 for centuries, and every time, it turned out to be a con — though sometimes
1860 the con artists were also conning themselves.
1861   </p><p>
1862     For generations, the advertising industry has been steadily improving its
1863 ability to sell advertising services to businesses while only making
1864 marginal gains in selling those businesses’ products to prospective
1865 customers. John Wanamaker’s lament that <span class="quote">« <span class="quote">50% of my advertising budget
1866 is wasted, I just don’t know which 50%</span> »</span> is a testament to the triumph
1867 of <span class="emphasis"><em>ad executives</em></span>, who successfully convinced Wanamaker
1868 that only half of the money he spent went to waste.
1869   </p><p>
1870     The tech industry has made enormous improvements in the science of
1871 convincing businesses that they’re good at advertising while their actual
1872 improvements to advertising — as opposed to targeting — have been pretty
1873 ho-hum. The vogue for machine learning — and the mystical invocation of
1874 <span class="quote">« <span class="quote">artificial intelligence</span> »</span> as a synonym for straightforward
1875 statistical inference techniques — has greatly boosted the efficacy of Big
1876 Tech’s sales pitch as marketers have exploited potential customers’ lack of
1877 technical sophistication to get away with breathtaking acts of overpromising
1878 and underdelivering.
1879   </p><p>
1880     It’s tempting to think that if businesses are willing to pour billions into
1881 a venture that the venture must be a good one. Yet there are plenty of times
1882 when this rule of thumb has led us astray.  For example, it’s virtually
1883 unheard of for managed investment funds to outperform simple index funds,
1884 and investors who put their money into the hands of expert money managers
1885 overwhelmingly fare worse than those who entrust their savings to index
1886 funds. But managed funds still account for the majority of the money
1887 invested in the markets, and they are patronized by some of the richest,
1888 most sophisticated investors in the world. Their vote of confidence in an
1889 underperforming sector is a parable about the role of luck in wealth
1890 accumulation, not a sign that managed funds are a good buy.
1891   </p><p>
1892     The claims of Big Tech’s mind-control system are full of tells that the
1893 enterprise is a con. For example, <a class="ulink" href="https://www.frontiersin.org/articles/10.3389/fpsyg.2020.01415/full" target="_top">the
1894 reliance on the <span class="quote">« <span class="quote">Big Five</span> »</span> personality traits</a> as a
1895 primary means of influencing people even though the <span class="quote">« <span class="quote">Big Five</span> »</span>
1896 theory is unsupported by any large-scale, peer-reviewed studies and is
1897 <a class="ulink" href="https://www.wired.com/story/the-noisy-fallacies-of-psychographic-targeting/" target="_top">mostly
1898 the realm of marketing hucksters and pop psych</a>.
1899   </p><p>
1900     Big Tech’s promotional materials also claim that their algorithms can
1901 accurately perform <span class="quote">« <span class="quote">sentiment analysis</span> »</span> or detect peoples’
1902 moods based on their <span class="quote">« <span class="quote">microexpressions,</span> »</span> but <a class="ulink" href="https://www.npr.org/2018/09/12/647040758/advertising-on-facebook-is-it-worth-it" target="_top">these
1903 are marketing claims, not scientific ones</a>. These methods are largely
1904 untested by independent scientific experts, and where they have been tested,
1905 they’ve been found sorely wanting.  Microexpressions are particularly
1906 suspect as the companies that specialize in training people to detect them
1907 <a class="ulink" href="https://theintercept.com/2017/02/08/tsas-own-files-show-doubtful-science-behind-its-behavior-screening-program/" target="_top">have
1908 been shown</a> to underperform relative to random chance.
1909   </p><p>
1910     Big Tech has been so good at marketing its own supposed superpowers that
1911 it’s easy to believe that they can market everything else with similar
1912 acumen, but it’s a mistake to believe the hype. Any statement a company
1913 makes about the quality of its products is clearly not impartial. The fact
1914 that we distrust all the things that Big Tech says about its data handling,
1915 compliance with privacy laws, etc., is only reasonable — but why on Earth
1916 would we treat Big Tech’s marketing literature as the gospel truth? Big Tech
1917 lies about just about <span class="emphasis"><em>everything</em></span>, including how well
1918 its machine-learning fueled persuasion systems work.
1919   </p><p>
1920     That skepticism should infuse all of our evaluations of Big Tech and its
1921 supposed abilities, including our perusal of its patents. Zuboff vests these
1922 patents with enormous significance, pointing out that Google claimed
1923 extensive new persuasion capabilities in <a class="ulink" href="https://patents.google.com/patent/US20050131762A1/en" target="_top">its patent
1924 filings</a>. These claims are doubly suspect: first, because they are so
1925 self-serving, and second, because the patent itself is so notoriously an
1926 invitation to exaggeration.
1927   </p><p>
1928     Patent applications take the form of a series of claims and range from broad
1929 to narrow. A typical patent starts out by claiming that its authors have
1930 invented a method or system for doing every conceivable thing that anyone
1931 might do, ever, with any tool or device. Then it narrows that claim in
1932 successive stages until we get to the actual <span class="quote">« <span class="quote">invention</span> »</span> that
1933 is the true subject of the patent.  The hope is that the patent examiner —
1934 who is almost certainly overworked and underinformed — will miss the fact
1935 that some or all of these claims are ridiculous, or at least suspect, and
1936 grant the patent’s broader claims. Patents for unpatentable things are still
1937 incredibly useful because they can be wielded against competitors who might
1938 license that patent or steer clear of its claims rather than endure the
1939 lengthy, expensive process of contesting it.
1940   </p><p>
1941     What’s more, software patents are routinely granted even though the filer
1942 doesn’t have any evidence that they can do the thing claimed by the
1943 patent. That is, you can patent an <span class="quote">« <span class="quote">invention</span> »</span> that you haven’t
1944 actually made and that you don’t know how to make.
1945   </p><p>
1946     With these considerations in hand, it becomes obvious that the fact that a
1947 Big Tech company has patented what it <span class="emphasis"><em>says</em></span> is an
1948 effective mind-control ray is largely irrelevant to whether Big Tech can in
1949 fact control our minds.
1950   </p><p>
1951     Big Tech collects our data for many reasons, including the diminishing
1952 returns on existing stores of data. But many tech companies also collect
1953 data out of a mistaken tech exceptionalist belief in the network effects of
1954 data. Network effects occur when each new user in a system increases its
1955 value. The classic example is fax machines: A single fax machine is of no
1956 use, two fax machines are of limited use, but every new fax machine that’s
1957 put to use after the first doubles the number of possible fax-to-fax links.
1958   </p><p>
1959     Data mined for predictive systems doesn’t necessarily produce these
1960 dividends. Think of Netflix: The predictive value of the data mined from a
1961 million English-speaking Netflix viewers is hardly improved by the addition
1962 of one more user’s viewing data. Most of the data Netflix acquires after
1963 that first minimum viable sample duplicates existing data and produces only
1964 minimal gains. Meanwhile, retraining models with new data gets progressively
1965 more expensive as the number of data points increases, and manual tasks like
1966 labeling and validating data do not get cheaper at scale.
1967   </p><p>
1968     Businesses pursue fads to the detriment of their profits all the time,
1969 especially when the businesses and their investors are not motivated by the
1970 prospect of becoming profitable but rather by the prospect of being acquired
1971 by a Big Tech giant or by having an IPO.  For these firms, ticking faddish
1972 boxes like <span class="quote">« <span class="quote">collects as much data as possible</span> »</span> might realize a
1973 bigger return on investment than <span class="quote">« <span class="quote">collects a business-appropriate
1974 quantity of data.</span> »</span>
1975   </p><p>
1976     This is another harm of tech exceptionalism: The belief that more data
1977 always produces more profits in the form of more insights that can be
1978 translated into better mind-control rays drives firms to over-collect and
1979 over-retain data beyond all rationality. And since the firms are behaving
1980 irrationally, a good number of them will go out of business and become ghost
1981 ships whose cargo holds are stuffed full of data that can harm people in
1982 myriad ways — but which no one is responsible for antey longer. Even if the
1983 companies don’t go under, the data they collect is maintained behind the
1984 minimum viable security — just enough security to keep the company viable
1985 while it waits to get bought out by a tech giant, an amount calculated to
1986 spend not one penny more than is necessary on protecting data.
1987   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="how-monopolies-not-mind-control-drive-surveillance-capitalism-the-snapchat-story"></a>How monopolies, not mind control, drive surveillance capitalism: The
1988 Snapchat story</h2></div></div></div><p>
1989     For the first decade of its existence, Facebook competed with the social
1990 media giants of the day (Myspace, Orkut, etc.) by presenting itself as the
1991 pro-privacy alternative. Indeed, Facebook justified its walled garden —
1992 which let users bring in data from the web but blocked web services like
1993 Google Search from indexing and caching Facebook pages — as a pro-privacy
1994 measure that protected users from the surveillance-happy winners of the
1995 social media wars like Myspace.
1996   </p><p>
1997     Despite frequent promises that it would never collect or analyze its users’
1998 data, Facebook periodically created initiatives that did just that, like the
1999 creepy, ham-fisted Beacon tool, which spied on you as you moved around the
2000 web and then added your online activities to your public timeline, allowing
2001 your friends to monitor your browsing habits. Beacon sparked a user
2002 revolt. Every time, Facebook backed off from its surveillance initiative,
2003 but not all the way; inevitably, the new Facebook would be more surveilling
2004 than the old Facebook, though not quite as surveilling as the intermediate
2005 Facebook following the launch of the new product or service.
2006   </p><p>
2007     The pace at which Facebook ramped up its surveillance efforts seems to have
2008 been set by Facebook’s competitive landscape. The more competitors Facebook
2009 had, the better it behaved. Every time a major competitor foundered,
2010 Facebook’s behavior <a class="ulink" href="https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3247362" target="_top">got
2011 markedly worse</a>.
2012   </p><p>
2013     All the while, Facebook was prodigiously acquiring companies, including a
2014 company called Onavo. Nominally, Onavo made a battery-monitoring mobile
2015 app. But the permissions that Onavo required were so expansive that the app
2016 was able to gather fine-grained telemetry on everything users did with their
2017 phones, including which apps they used and how they were using them.
2018   </p><p>
2019     Through Onavo, Facebook discovered that it was losing market share to
2020 Snapchat, an app that — like Facebook a decade before — billed itself as the
2021 pro-privacy alternative to the status quo. Through Onavo, Facebook was able
2022 to mine data from the devices of Snapchat users, including both current and
2023 former Snapchat users. This spurred Facebook to acquire Instagram — some
2024 features of which competed with Snapchat — and then allowed Facebook to
2025 fine-tune Instagram’s features and sales pitch to erode Snapchat’s gains and
2026 ensure that Facebook would not have to face the kinds of competitive
2027 pressures it had earlier inflicted on Myspace and Orkut.
2028   </p><p>
2029     The story of how Facebook crushed Snapchat reveals the relationship between
2030 monopoly and surveillance capitalism. Facebook combined surveillance with
2031 lax antitrust enforcement to spot the competitive threat of Snapchat on its
2032 horizon and then take decisive action against it. Facebook’s surveillance
2033 capitalism let it avert competitive pressure with anti-competitive
2034 tactics. Facebook users still want privacy — Facebook hasn’t used
2035 surveillance to brainwash them out of it — but they can’t get it because
2036 Facebook’s surveillance lets it destroy any hope of a rival service emerging
2037 that competes on privacy features.
2038   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="a-monopoly-over-your-friends"></a>A monopoly over your friends</h2></div></div></div><p>
2039     A decentralization movement has tried to erode the dominance of Facebook and
2040 other Big Tech companies by fielding <span class="quote">« <span class="quote">indieweb</span> »</span> alternatives —
2041 Mastodon as a Twitter alternative, Diaspora as a Facebook alternative,
2042 etc. — but these efforts have failed to attain any kind of liftoff.
2043   </p><p>
2044     Fundamentally, each of these services is hamstrung by the same problem:
2045 Every potential user for a Facebook or Twitter alternative has to convince
2046 all their friends to follow them to a decentralized web alternative in order
2047 to continue to realize the benefit of social media. For many of us, the only
2048 reason to have a Facebook account is that our friends have Facebook
2049 accounts, and the reason they have Facebook accounts is that
2050 <span class="emphasis"><em>we</em></span> have Facebook accounts.
2051   </p><p>
2052     All of this has conspired to make Facebook — and other dominant platforms —
2053 into <span class="quote">« <span class="quote">kill zones</span> »</span> that investors will not fund new entrants
2054 for.
2055   </p><p>
2056     And yet, all of today’s tech giants came into existence despite the
2057 entrenched advantage of the companies that came before them. To understand
2058 how that happened, you have to understand both interoperability and
2059 adversarial interoperability.
2060   </p><div class="blockquote"><blockquote class="blockquote"><p>
2061       The hard problem of our species is coordination.
2062     </p></blockquote></div><p>
2063     <span class="quote">« <span class="quote">Interoperability</span> »</span> is the ability of two technologies to work
2064 with one another: Anyone can make an LP that will play on any record player,
2065 anyone can make a filter you can install in your stove’s extractor fan,
2066 anyone can make gasoline for your car, anyone can make a USB phone charger
2067 that fits in your car’s cigarette lighter receptacle, anyone can make a
2068 light bulb that works in your light socket, anyone can make bread that will
2069 toast in your toaster.
2070   </p><p>
2071     Interoperability is often a source of innovation and consumer benefit: Apple
2072 made the first commercially successful PC, but millions of independent
2073 software vendors made interoperable programs that ran on the Apple II
2074 Plus. The simple analog antenna inputs on the back of TVs first allowed
2075 cable operators to connect directly to TVs, then they allowed game console
2076 companies and then personal computer companies to use standard televisions
2077 as displays. Standard RJ-11 telephone jacks allowed for the production of
2078 phones from a variety of vendors in a variety of forms, from the free
2079 football-shaped phone that came with a <span class="emphasis"><em>Sports
2080 Illustrated</em></span> subscription to business phones with speakers, hold
2081 functions, and so on and then answering machines and finally modems, paving
2082 the way for the internet revolution.
2083   </p><p>
2084     <span class="quote">« <span class="quote">Interoperability</span> »</span> is often used interchangeably with
2085 <span class="quote">« <span class="quote">standardization,</span> »</span> which is the process when manufacturers and
2086 other stakeholders hammer out a set of agreed-upon rules for implementing a
2087 technology, such as the electrical plug on your wall, the CAN bus used by
2088 your car’s computer systems, or the HTML instructions that your browser
2089 interprets.
2090   </p><p>
2091     But interoperability doesn’t require standardization — indeed,
2092 standardization often proceeds from the chaos of ad hoc interoperability
2093 measures. The inventor of the cigarette-lighter USB charger didn’t need to
2094 get permission from car manufacturers or even the manufacturers of the
2095 dashboard lighter subcomponent. The automakers didn’t take any
2096 countermeasures to prevent the use of these aftermarket accessories by their
2097 customers, but they also didn’t do anything to make life easier for the
2098 chargers’ manufacturers. This is a kind of <span class="quote">« <span class="quote">neutral
2099 interoperability.</span> »</span>
2100   </p><p>
2101     Beyond neutral interoperability, there is <span class="quote">« <span class="quote">adversarial
2102 interoperability.</span> »</span> That’s when a manufacturer makes a product that
2103 interoperates with another manufacturer’s product <span class="emphasis"><em>despite the
2104 second manufacturer’s objections</em></span> and <span class="emphasis"><em>even if that means
2105 bypassing a security system designed to prevent interoperability</em></span>.
2106   </p><p>
2107     Probably the most familiar form of adversarial interoperability is
2108 third-party printer ink. Printer manufacturers claim that they sell printers
2109 below cost and that the only way they can recoup the losses they incur is by
2110 charging high markups on ink. To prevent the owners of printers from buying
2111 ink elsewhere, the printer companies deploy a suite of anti-customer
2112 security systems that detect and reject both refilled and third-party
2113 cartridges.
2114   </p><p>
2115     Owners of printers take the position that HP and Epson and Brother are not
2116 charities and that customers for their wares have no obligation to help them
2117 survive, and so if the companies choose to sell their products at a loss,
2118 that’s their foolish choice and their consequences to live with. Likewise,
2119 competitors who make ink or refill kits observe that they don’t owe printer
2120 companies anything, and their erosion of printer companies’ margins are the
2121 printer companies’ problems, not their competitors’. After all, the printer
2122 companies shed no tears when they drive a refiller out of business, so why
2123 should the refillers concern themselves with the economic fortunes of the
2124 printer companies?
2125   </p><p>
2126     Adversarial interoperability has played an outsized role in the history of
2127 the tech industry: from the founding of the <span class="quote">« <span class="quote">alt.*</span> »</span> Usenet
2128 hierarchy (which was started against the wishes of Usenet’s maintainers and
2129 which grew to be bigger than all of Usenet combined)  to the browser wars
2130 (when Netscape and Microsoft devoted massive engineering efforts to making
2131 their browsers incompatible with the other’s special commands and
2132 peccadilloes) to Facebook (whose success was built in part by helping its
2133 new users stay in touch with friends they’d left behind on Myspace because
2134 Facebook supplied them with a tool that scraped waiting messages from
2135 Myspace and imported them into Facebook, effectively creating an
2136 Facebook-based Myspace reader).
2137   </p><p>
2138     Today, incumbency is seen as an unassailable advantage. Facebook is where
2139 all of your friends are, so no one can start a Facebook competitor. But
2140 adversarial compatibility reverses the competitive advantage: If you were
2141 allowed to compete with Facebook by providing a tool that imported all your
2142 users’ waiting Facebook messages into an environment that competed on lines
2143 that Facebook couldn’t cross, like eliminating surveillance and ads, then
2144 Facebook would be at a huge disadvantage. It would have assembled all
2145 possible ex-Facebook users into a single, easy-to-find service; it would
2146 have educated them on how a Facebook-like service worked and what its
2147 potential benefits were; and it would have provided an easy means for
2148 disgruntled Facebook users to tell their friends where they might expect
2149 better treatment.
2150   </p><p>
2151     Adversarial interoperability was once the norm and a key contributor to the
2152 dynamic, vibrant tech scene, but now it is stuck behind a thicket of laws
2153 and regulations that add legal risks to the tried-and-true tactics of
2154 adversarial interoperability. New rules and new interpretations of existing
2155 rules mean that a would-be adversarial interoperator needs to steer clear of
2156 claims under copyright, terms of service, trade secrecy, tortious
2157 interference, and patent.
2158   </p><p>
2159     In the absence of a competitive market, lawmakers have resorted to assigning
2160 expensive, state-like duties to Big Tech firms, such as automatically
2161 filtering user contributions for copyright infringement or terrorist and
2162 extremist content or detecting and preventing harassment in real time or
2163 controlling access to sexual material.
2164   </p><p>
2165     These measures put a floor under how small we can make Big Tech because only
2166 the very largest companies can afford the humans and automated filters
2167 needed to perform these duties.
2168   </p><p>
2169     But that’s not the only way in which making platforms responsible for
2170 policing their users undermines competition. A platform that is expected to
2171 police its users’ conduct must prevent many vital adversarial
2172 interoperability techniques lest these subvert its policing measures. For
2173 example, if someone using a Twitter replacement like Mastodon is able to
2174 push messages into Twitter and read messages out of Twitter, they could
2175 avoid being caught by automated systems that detect and prevent harassment
2176 (such as systems that use the timing of messages or IP-based rules to make
2177 guesses about whether someone is a harasser).
2178   </p><p>
2179     To the extent that we are willing to let Big Tech police itself — rather
2180 than making Big Tech small enough that users can leave bad platforms for
2181 better ones and small enough that a regulation that simply puts a platform
2182 out of business will not destroy billions of users’ access to their
2183 communities and data — we build the case that Big Tech should be able to
2184 block its competitors and make it easier for Big Tech to demand legal
2185 enforcement tools to ban and punish attempts at adversarial
2186 interoperability.
2187   </p><p>
2188     Ultimately, we can try to fix Big Tech by making it responsible for bad acts
2189 by its users, or we can try to fix the internet by cutting Big Tech down to
2190 size. But we can’t do both. To replace today’s giant products with
2191 pluralistic protocols, we need to clear the legal thicket that prevents
2192 adversarial interoperability so that tomorrow’s nimble, personal,
2193 small-scale products can federate themselves with giants like Facebook,
2194 allowing the users who’ve left to continue to communicate with users who
2195 haven’t left yet, reaching tendrils over Facebook’s garden wall that
2196 Facebook’s trapped users can use to scale the walls and escape to the
2197 global, open web.
2198   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="fake-news-is-an-epistemological-crisis"></a>Fake news is an epistemological crisis</h2></div></div></div><p>
2199     Tech is not the only industry that has undergone massive concentration since
2200 the Reagan era. Virtually every major industry — from oil to newspapers to
2201 meatpacking to sea freight to eyewear to online pornography — has become a
2202 clubby oligarchy that just a few players dominate.
2203   </p><p>
2204     At the same time, every industry has become something of a tech industry as
2205 general-purpose computers and general-purpose networks and the promise of
2206 efficiencies through data-driven analysis infuse every device, process, and
2207 firm with tech.
2208   </p><p>
2209     This phenomenon of industrial concentration is part of a wider story about
2210 wealth concentration overall as a smaller and smaller number of people own
2211 more and more of our world. This concentration of both wealth and industries
2212 means that our political outcomes are increasingly beholden to the parochial
2213 interests of the people and companies with all the money.
2214   </p><p>
2215     That means that whenever a regulator asks a question with an obvious,
2216 empirical answer (<span class="quote">« <span class="quote">Are humans causing climate change?</span> »</span> or
2217 <span class="quote">« <span class="quote">Should we let companies conduct commercial mass surveillance?</span> »</span>
2218 or <span class="quote">« <span class="quote">Does society benefit from allowing network neutrality
2219 violations?</span> »</span>), the answer that comes out is only correct if that
2220 correctness meets with the approval of rich people and the industries that
2221 made them so wealthy.
2222   </p><p>
2223     Rich people have always played an outsized role in politics and more so
2224 since the Supreme Court’s <span class="emphasis"><em>Citizens United</em></span> decision
2225 eliminated key controls over political spending. Widening inequality and
2226 wealth concentration means that the very richest people are now a lot richer
2227 and can afford to spend a lot more money on political projects than ever
2228 before. Think of the Koch brothers or George Soros or Bill Gates.
2229   </p><p>
2230     But the policy distortions of rich individuals pale in comparison to the
2231 policy distortions that concentrated industries are capable of.  The
2232 companies in highly concentrated industries are much more profitable than
2233 companies in competitive industries — no competition means not having to
2234 reduce prices or improve quality to win customers — leaving them with bigger
2235 capital surpluses to spend on lobbying.
2236   </p><p>
2237     Concentrated industries also find it easier to collaborate on policy
2238 objectives than competitive ones. When all the top execs from your industry
2239 can fit around a single boardroom table, they often do. And
2240 <span class="emphasis"><em>when</em></span> they do, they can forge a consensus position on
2241 regulation.
2242   </p><p>
2243     Rising through the ranks in a concentrated industry generally means working
2244 at two or three of the big companies. When there are only relatively few
2245 companies in a given industry, each company has a more ossified executive
2246 rank, leaving ambitious execs with fewer paths to higher positions unless
2247 they are recruited to a rival. This means that the top execs in concentrated
2248 industries are likely to have been colleagues at some point and socialize in
2249 the same circles — connected through social ties or, say, serving as
2250 trustees for each others’ estates. These tight social bonds foster a
2251 collegial, rather than competitive, attitude.
2252   </p><p>
2253     Highly concentrated industries also present a regulatory conundrum.  When an
2254 industry is dominated by just four or five companies, the only people who
2255 are likely to truly understand the industry’s practices are its veteran
2256 executives. This means that top regulators are often former execs of the
2257 companies they are supposed to be regulating. These turns in government are
2258 often tacitly understood to be leaves of absence from industry, with former
2259 employers welcoming their erstwhile watchdogs back into their executive
2260 ranks once their terms have expired.
2261   </p><p>
2262     All this is to say that the tight social bonds, small number of firms, and
2263 regulatory capture of concentrated industries give the companies that
2264 comprise them the power to dictate many, if not all, of the regulations that
2265 bind them.
2266   </p><p>
2267     This is increasingly obvious. Whether it’s payday lenders <a class="ulink" href="https://www.washingtonpost.com/business/2019/02/25/how-payday-lending-industry-insider-tilted-academic-research-its-favor/" target="_top">winning
2268 the right to practice predatory lending</a> or Apple <a class="ulink" href="https://www.vice.com/en_us/article/mgxayp/source-apple-will-fight-right-to-repair-legislation" target="_top">winning
2269 the right to decide who can fix your phone</a> or Google and Facebook
2270 winning the right to breach your private data without suffering meaningful
2271 consequences or victories for pipeline companies or impunity for opioid
2272 manufacturers or massive tax subsidies for incredibly profitable dominant
2273 businesses, it’s increasingly apparent that many of our official,
2274 evidence-based truth-seeking processes are, in fact, auctions for sale to
2275 the highest bidder.
2276   </p><p>
2277     It’s really impossible to overstate what a terrifying prospect this is. We
2278 live in an incredibly high-tech society, and none of us could acquire the
2279 expertise to evaluate every technological proposition that stands between us
2280 and our untimely, horrible deaths. You might devote your life to acquiring
2281 the media literacy to distinguish good scientific journals from corrupt
2282 pay-for-play lookalikes and the statistical literacy to evaluate the quality
2283 of the analysis in the journals as well as the microbiology and epidemiology
2284 knowledge to determine whether you can trust claims about the safety of
2285 vaccines — but that would still leave you unqualified to judge whether the
2286 wiring in your home will give you a lethal shock <span class="emphasis"><em>and</em></span>
2287 whether your car’s brakes’ software will cause them to fail unpredictably
2288 <span class="emphasis"><em>and</em></span> whether the hygiene standards at your butcher are
2289 sufficient to keep you from dying after you finish your dinner.
2290   </p><p>
2291     In a world as complex as this one, we have to defer to authorities, and we
2292 keep them honest by making those authorities accountable to us and binding
2293 them with rules to prevent conflicts of interest. We can’t possibly acquire
2294 the expertise to adjudicate conflicting claims about the best way to make
2295 the world safe and prosperous, but we <span class="emphasis"><em>can</em></span> determine
2296 whether the adjudication process itself is trustworthy.
2297   </p><p>
2298     Right now, it’s obviously not.
2299   </p><p>
2300     The past 40 years of rising inequality and industry concentration, together
2301 with increasingly weak accountability and transparency for expert agencies,
2302 has created an increasingly urgent sense of impending doom, the sense that
2303 there are vast conspiracies afoot that operate with tacit official approval
2304 despite the likelihood they are working to better themselves by ruining the
2305 rest of us.
2306   </p><p>
2307     For example, it’s been decades since Exxon’s own scientists concluded that
2308 its products would render the Earth uninhabitable by humans. And yet those
2309 decades were lost to us, in large part because Exxon lobbied governments and
2310 sowed doubt about the dangers of its products and did so with the
2311 cooperation of many public officials.  When the survival of you and everyone
2312 you love is threatened by conspiracies, it’s not unreasonable to start
2313 questioning the things you think you know in an attempt to determine whether
2314 they, too, are the outcome of another conspiracy.
2315   </p><p>
2316     The collapse of the credibility of our systems for divining and upholding
2317 truths has left us in a state of epistemological chaos.  Once, most of us
2318 might have assumed that the system was working and that our regulations
2319 reflected our best understanding of the empirical truths of the world as
2320 they were best understood — now we have to find our own experts to help us
2321 sort the true from the false.
2322   </p><p>
2323     If you’re like me, you probably believe that vaccines are safe, but you
2324 (like me) probably also can’t explain the microbiology or statistics. Few of
2325 us have the math skills to review the literature on vaccine safety and
2326 describe why their statistical reasoning is sound. Likewise, few of us can
2327 review the stats in the (now discredited) literature on opioid safety and
2328 explain how those stats were manipulated. Both vaccines and opioids were
2329 embraced by medical authorities, after all, and one is safe while the other
2330 could ruin your life. You’re left with a kind of inchoate constellation of
2331 rules of thumb about which experts you trust to fact-check controversial
2332 claims and then to explain how all those respectable doctors with their
2333 peer-reviewed research on opioid safety <span class="emphasis"><em>were</em></span> an
2334 aberration and then how you know that the doctors writing about vaccine
2335 safety are <span class="emphasis"><em>not</em></span> an aberration.
2336   </p><p>
2337     I’m 100% certain that vaccinating is safe and effective, but I’m also at
2338 something of a loss to explain exactly, <span class="emphasis"><em>precisely,</em></span> why
2339 I believe this, given all the corruption I know about and the many times the
2340 stamp of certainty has turned out to be a parochial lie told to further
2341 enrich the super rich.
2342   </p><p>
2343     Fake news — conspiracy theories, racist ideologies, scientific denialism —
2344 has always been with us. What’s changed today is not the mix of ideas in the
2345 public discourse but the popularity of the worst ideas in that
2346 mix. Conspiracy and denial have skyrocketed in lockstep with the growth of
2347 Big Inequality, which has also tracked the rise of Big Tech and Big Pharma
2348 and Big Wrestling and Big Car and Big Movie Theater and Big Everything Else.
2349   </p><p>
2350     No one can say for certain why this has happened, but the two dominant camps
2351 are idealism (the belief that the people who argue for these conspiracies
2352 have gotten better at explaining them, maybe with the help of
2353 machine-learning tools) or materialism (the ideas have become more
2354 attractive because of material conditions in the world).
2355   </p><p>
2356     I’m a materialist. I’ve been exposed to the arguments of conspiracy
2357 theorists all my life, and I have not experienced any qualitative leap in
2358 the quality of those arguments.
2359   </p><p>
2360     The major difference is in the world, not the arguments. In a time where
2361 actual conspiracies are commonplace, conspiracy theories acquire a ring of
2362 plausibility.
2363   </p><p>
2364     We have always had disagreements about what’s true, but today, we have a
2365 disagreement over how we know whether something is true. This is an
2366 epistemological crisis, not a crisis over belief. It’s a crisis over the
2367 credibility of our truth-seeking exercises, from scientific journals (in an
2368 era where the biggest journal publishers have been caught producing
2369 pay-to-play journals for junk science) to regulations (in an era where
2370 regulators are routinely cycling in and out of business) to education (in an
2371 era where universities are dependent on corporate donations to keep their
2372 lights on).
2373   </p><p>
2374     Targeting — surveillance capitalism — makes it easier to find people who are
2375 undergoing this epistemological crisis, but it doesn’t create the
2376 crisis. For that, you need to look to corruption.
2377   </p><p>
2378     And, conveniently enough, it’s corruption that allows surveillance
2379 capitalism to grow by dismantling monopoly protections, by permitting
2380 reckless collection and retention of personal data, by allowing ads to be
2381 targeted in secret, and by foreclosing on the possibility of going somewhere
2382 else where you might continue to enjoy your friends without subjecting
2383 yourself to commercial surveillance.
2384   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="tech-is-different"></a>Tech is different</h2></div></div></div><p>
2385     I reject both iterations of technological exceptionalism. I reject the idea
2386 that tech is uniquely terrible and led by people who are greedier or worse
2387 than the leaders of other industries, and I reject the idea that tech is so
2388 good — or so intrinsically prone to concentration — that it can’t be blamed
2389 for its present-day monopolistic status.
2390   </p><p>
2391     I think tech is just another industry, albeit one that grew up in the
2392 absence of real monopoly constraints. It may have been first, but it isn’t
2393 the worst nor will it be the last.
2394   </p><p>
2395     But there’s one way in which I <span class="emphasis"><em>am</em></span> a tech
2396 exceptionalist. I believe that online tools are the key to overcoming
2397 problems that are much more urgent than tech monopolization: climate change,
2398 inequality, misogyny, and discrimination on the basis of race, gender
2399 identity, and other factors. The internet is how we will recruit people to
2400 fight those fights, and how we will coordinate their labor. Tech is not a
2401 substitute for democratic accountability, the rule of law, fairness, or
2402 stability — but it’s a means to achieve these things.
2403   </p><p>
2404     The hard problem of our species is coordination. Everything from climate
2405 change to social change to running a business to making a family work can be
2406 viewed as a collective action problem.
2407   </p><p>
2408     The internet makes it easier than at any time before to find people who want
2409 to work on a project with you — hence the success of free and open-source
2410 software, crowdfunding, and racist terror groups — and easier than ever to
2411 coordinate the work you do.
2412   </p><p>
2413     The internet and the computers we connect to it also possess an exceptional
2414 quality: general-purposeness. The internet is designed to allow any two
2415 parties to communicate any data, using any protocol, without permission from
2416 anyone else. The only production design we have for computers is the
2417 general-purpose, <span class="quote">« <span class="quote">Turing complete</span> »</span> computer that can run every
2418 program we can express in symbolic logic.
2419   </p><p>
2420     This means that every time someone with a special communications need
2421 invests in infrastructure and techniques to make the internet faster,
2422 cheaper, and more robust, this benefit redounds to everyone else who is
2423 using the internet to communicate. And this also means that every time
2424 someone with a special computing need invests to make computers faster,
2425 cheaper, and more robust, every other computing application is a potential
2426 beneficiary of this work.
2427   </p><p>
2428     For these reasons, every type of communication is gradually absorbed into
2429 the internet, and every type of device — from airplanes to pacemakers —
2430 eventually becomes a computer in a fancy case.
2431   </p><p>
2432     While these considerations don’t preclude regulating networks and computers,
2433 they do call for gravitas and caution when doing so because changes to
2434 regulatory frameworks could ripple out to have unintended consequences in
2435 many, many other domains.
2436   </p><p>
2437     The upshot of this is that our best hope of solving the big coordination
2438 problems — climate change, inequality, etc. — is with free, fair, and open
2439 tech. Our best hope of keeping tech free, fair, and open is to exercise
2440 caution in how we regulate tech and to attend closely to the ways in which
2441 interventions to solve one problem might create problems in other domains.
2442   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="ownership-of-facts"></a>Ownership of facts</h2></div></div></div><p>
2443     Big Tech has a funny relationship with information. When you’re generating
2444 information — anything from the location data streaming off your mobile
2445 device to the private messages you send to friends on a social network — it
2446 claims the rights to make unlimited use of that data.
2447   </p><p>
2448     But when you have the audacity to turn the tables — to use a tool that
2449 blocks ads or slurps your waiting updates out of a social network and puts
2450 them in another app that lets you set your own priorities and suggestions or
2451 crawls their system to allow you to start a rival business — they claim that
2452 you’re stealing from them.
2453   </p><p>
2454     The thing is, information is a very bad fit for any kind of private property
2455 regime. Property rights are useful for establishing markets that can lead to
2456 the effective development of fallow assets. These markets depend on clear
2457 titles to ensure that the things being bought and sold in them can, in fact,
2458 be bought and sold.
2459   </p><p>
2460     Information rarely has such a clear title. Take phone numbers: There’s
2461 clearly something going wrong when Facebook slurps up millions of users’
2462 address books and uses the phone numbers it finds in them to plot out social
2463 graphs and fill in missing information about other users.
2464   </p><p>
2465     But the phone numbers Facebook nonconsensually acquires in this transaction
2466 are not the <span class="quote">« <span class="quote">property</span> »</span> of the users they’re taken from nor do
2467 they belong to the people whose phones ring when you dial those numbers. The
2468 numbers are mere integers, 10 digits in the U.S.  and Canada, and they
2469 appear in millions of places, including somewhere deep in pi as well as
2470 numerous other contexts. Giving people ownership titles to integers is an
2471 obviously terrible idea.
2472   </p><p>
2473     Likewise for the facts that Facebook and other commercial surveillance
2474 operators acquire about us, like that we are the children of our parents or
2475 the parents to our children or that we had a conversation with someone else
2476 or went to a public place.  These data points can’t be property in the sense
2477 that your house or your shirt is your property because the title to them is
2478 intrinsically muddy: Does your mom own the fact that she is your mother? Do
2479 you? Do both of you? What about your dad — does he own this fact too, or
2480 does he have to license the fact from you (or your mom or both of you) in
2481 order to use this fact? What about the hundreds or thousands of other people
2482 who know these facts?
2483   </p><p>
2484     If you go to a Black Lives Matter demonstration, do the other demonstrators
2485 need your permission to post their photos from the event? The online fights
2486 over <a class="ulink" href="https://www.wired.com/story/how-to-take-photos-at-protests/" target="_top">when and
2487 how to post photos from demonstrations</a> reveal a nuanced, complex
2488 issue that cannot be easily hand-waved away by giving one party a property
2489 right that everyone else in the mix has to respect.
2490   </p><p>
2491     The fact that information isn’t a good fit with property and markets doesn’t
2492 mean that it’s not valuable. Babies aren’t property, but they’re inarguably
2493 valuable. In fact, we have a whole set of rules just for babies as well as a
2494 subset of those rules that apply to humans more generally. Someone who
2495 argues that babies won’t be truly valuable until they can be bought and sold
2496 like loaves of bread would be instantly and rightfully condemned as a
2497 monster.
2498   </p><p>
2499     It’s tempting to reach for the property hammer when Big Tech treats your
2500 information like a nail — not least because Big Tech are such prolific
2501 abusers of property hammers when it comes to <span class="emphasis"><em>their</em></span>
2502 information. But this is a mistake. If we allow markets to dictate the use
2503 of our information, then we’ll find that we’re sellers in a buyers’ market
2504 where the Big Tech monopolies set a price for our data that is so low as to
2505 be insignificant or, more likely, set at a nonnegotiable price of zero in a
2506 click-through agreement that you don’t have the opportunity to modify.
2507   </p><p>
2508     Meanwhile, establishing property rights over information will create
2509 insurmountable barriers to independent data processing. Imagine that we
2510 require a license to be negotiated when a translated document is compared
2511 with its original, something Google has done and continues to do billions of
2512 times to train its automated language translation tools. Google can afford
2513 this, but independent third parties cannot.  Google can staff a clearances
2514 department to negotiate one-time payments to the likes of the EU (one of the
2515 major repositories of translated documents) while independent watchdogs
2516 wanting to verify that the translations are well-prepared, or to root out
2517 bias in translations, will find themselves needing a staffed-up legal
2518 department and millions for licenses before they can even get started.
2519   </p><p>
2520     The same goes for things like search indexes of the web or photos of
2521 peoples’ houses, which have become contentious thanks to Google’s Street
2522 View project. Whatever problems may exist with Google’s photographing of
2523 street scenes, resolving them by letting people decide who can take pictures
2524 of the facades of their homes from a public street will surely create even
2525 worse ones. Think of how street photography is important for newsgathering —
2526 including informal newsgathering, like photographing abuses of authority —
2527 and how being able to document housing and street life are important for
2528 contesting eminent domain, advocating for social aid, reporting planning and
2529 zoning violations, documenting discriminatory and unequal living conditions,
2530 and more.
2531   </p><p>
2532     The ownership of facts is antithetical to many kinds of human progress. It’s
2533 hard to imagine a rule that limits Big Tech’s exploitation of our collective
2534 labors without inadvertently banning people from gathering data on online
2535 harassment or compiling indexes of changes in language or simply
2536 investigating how the platforms are shaping our discourse — all of which
2537 require scraping data that other people have created and subjecting it to
2538 scrutiny and analysis.
2539   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="persuasion-works-slowly"></a>Persuasion works… slowly</h2></div></div></div><p>
2540     The platforms may oversell their ability to persuade people, but obviously,
2541 persuasion works sometimes. Whether it’s the private realm that LGBTQ people
2542 used to recruit allies and normalize sexual diversity or the decadeslong
2543 project to convince people that markets are the only efficient way to solve
2544 complicated resource allocation problems, it’s clear that our societal
2545 attitudes <span class="emphasis"><em>can</em></span> change.
2546   </p><p>
2547     The project of shifting societal attitudes is a game of inches and
2548 years. For centuries, svengalis have purported to be able to accelerate this
2549 process, but even the most brutal forms of propaganda have struggled to make
2550 permanent changes. Joseph Goebbels was able to subject Germans to daily,
2551 mandatory, hourslong radio broadcasts, to round up and torture and murder
2552 dissidents, and to seize full control over their children’s education while
2553 banning any literature, broadcasts, or films that did not comport with his
2554 worldview.
2555   </p><p>
2556     Yet, after 12 years of terror, once the war ended, Nazi ideology was largely
2557 discredited in both East and West Germany, and a program of national truth
2558 and reconciliation was put in its place. Racism and authoritarianism were
2559 never fully abolished in Germany, but neither were the majority of Germans
2560 irrevocably convinced of Nazism — and the rise of racist authoritarianism in
2561 Germany today tells us that the liberal attitudes that replaced Nazism were
2562 no more permanent than Nazism itself.
2563   </p><p>
2564     Racism and authoritarianism have also always been with us. Anyone who’s
2565 reviewed the kind of messages and arguments that racists put forward today
2566 would be hard-pressed to say that they have gotten better at presenting
2567 their ideas. The same pseudoscience, appeals to fear, and circular logic
2568 that racists presented in the 1980s, when the cause of white supremacy was
2569 on the wane, are to be found in the communications of leading white
2570 nationalists today.
2571   </p><p>
2572     If racists haven’t gotten more convincing in the past decade, then how is it
2573 that more people were convinced to be openly racist at that time? I believe
2574 that the answer lies in the material world, not the world of ideas. The
2575 ideas haven’t gotten more convincing, but people have become more
2576 afraid. Afraid that the state can’t be trusted to act as an honest broker in
2577 life-or-death decisions, from those regarding the management of the economy
2578 to the regulation of painkillers to the rules for handling private
2579 information. Afraid that the world has become a game of musical chairs in
2580 which the chairs are being taken away at a never-before-seen rate. Afraid
2581 that justice for others will come at their expense. Monopolism isn’t the
2582 cause of these fears, but the inequality and material desperation and policy
2583 malpractice that monopolism contributes to is a significant contributor to
2584 these conditions. Inequality creates the conditions for both conspiracies
2585 and violent racist ideologies, and then surveillance capitalism lets
2586 opportunists target the fearful and the conspiracy-minded.
2587   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="paying-wont-help"></a>Paying won’t help</h2></div></div></div><p>
2588     As the old saw goes, <span class="quote">« <span class="quote">If you’re not paying for the product, you’re the
2589 product.</span> »</span>
2590   </p><p>
2591     It’s a commonplace belief today that the advent of free, ad-supported media
2592 was the original sin of surveillance capitalism.  The reasoning is that the
2593 companies that charged for access couldn’t <span class="quote">« <span class="quote">compete with free</span> »</span>
2594 and so they were driven out of business. Their ad-supported competitors,
2595 meanwhile, declared open season on their users’ data in a bid to improve
2596 their ad targeting and make more money and then resorted to the most
2597 sensationalist tactics to generate clicks on those ads. If only we’d pay for
2598 media again, we’d have a better, more responsible, more sober discourse that
2599 would be better for democracy.
2600   </p><p>
2601     But the degradation of news products long precedes the advent of
2602 ad-supported online news. Long before newspapers were online, lax antitrust
2603 enforcement had opened the door for unprecedented waves of consolidation and
2604 roll-ups in newsrooms. Rival newspapers were merged, reporters and ad sales
2605 staff were laid off, physical plants were sold and leased back, leaving the
2606 companies loaded up with debt through leveraged buyouts and subsequent
2607 profit-taking by the new owners. In other words, it wasn’t merely shifts in
2608 the classified advertising market, which was long held to be the primary
2609 driver in the decline of the traditional newsroom, that made news companies
2610 unable to adapt to the internet — it was monopolism.
2611   </p><p>
2612     Then, as news companies <span class="emphasis"><em>did</em></span> come online, the ad
2613 revenues they commanded dropped even as the number of internet users (and
2614 thus potential online readers) increased. That shift was a function of
2615 consolidation in the ad sales market, with Google and Facebook emerging as
2616 duopolists who made more money every year from advertising while paying less
2617 and less of it to the publishers whose work the ads appeared
2618 alongside. Monopolism created a buyer’s market for ad inventory with
2619 Facebook and Google acting as gatekeepers.
2620   </p><p>
2621     Paid services continue to exist alongside free ones, and often it is these
2622 paid services — anxious to prevent people from bypassing their paywalls or
2623 sharing paid media with freeloaders — that exert the most control over their
2624 customers. Apple’s iTunes and App Stores are paid services, but to maximize
2625 their profitability, Apple has to lock its platforms so that third parties
2626 can’t make compatible software without permission. These locks allow the
2627 company to exercise both editorial control (enabling it to exclude <a class="ulink" href="https://ncac.org/news/blog/does-apples-strict-app-store-content-policy-limit-freedom-of-expression" target="_top">controversial
2628 political material</a>) and technological control, including control
2629 over who can repair the devices it makes. If we’re worried that ad-supported
2630 products deprive people of their right to self-determination by using
2631 persuasion techniques to nudge their purchase decisions a few degrees in one
2632 direction or the other, then the near-total control a single company holds
2633 over the decision of who gets to sell you software, parts, and service for
2634 your iPhone should have us very worried indeed.
2635   </p><p>
2636     We shouldn’t just be concerned about payment and control: The idea that
2637 paying will improve discourse is also dangerously wrong. The poor success
2638 rate of targeted advertising means that the platforms have to incentivize
2639 you to <span class="quote">« <span class="quote">engage</span> »</span> with posts at extremely high levels to generate
2640 enough pageviews to safeguard their profits. As discussed earlier, to
2641 increase engagement, platforms like Facebook use machine learning to guess
2642 which messages will be most inflammatory and make a point of shoving those
2643 into your eyeballs at every turn so that you will hate-click and argue with
2644 people.
2645   </p><p>
2646     Perhaps paying would fix this, the reasoning goes. If platforms could be
2647 economically viable even if you stopped clicking on them once your
2648 intellectual and social curiosity had been slaked, then they would have no
2649 reason to algorithmically enrage you to get more clicks out of you, right?
2650   </p><p>
2651     There may be something to that argument, but it still ignores the wider
2652 economic and political context of the platforms and the world that allowed
2653 them to grow so dominant.
2654   </p><p>
2655     Platforms are world-spanning and all-encompassing because they are
2656 monopolies, and they are monopolies because we have gutted our most
2657 important and reliable anti-monopoly rules. Antitrust was neutered as a key
2658 part of the project to make the wealthy wealthier, and that project has
2659 worked. The vast majority of people on Earth have a negative net worth, and
2660 even the dwindling middle class is in a precarious state, undersaved for
2661 retirement, underinsured for medical disasters, and undersecured against
2662 climate and technology shocks.
2663   </p><p>
2664     In this wildly unequal world, paying doesn’t improve the discourse; it
2665 simply prices discourse out of the range of the majority of people. Paying
2666 for the product is dandy, if you can afford it.
2667   </p><p>
2668     If you think today’s filter bubbles are a problem for our discourse, imagine
2669 what they’d be like if rich people inhabited free-flowing Athenian
2670 marketplaces of ideas where you have to pay for admission while everyone
2671 else lives in online spaces that are subsidized by wealthy benefactors who
2672 relish the chance to establish conversational spaces where the <span class="quote">« <span class="quote">house
2673 rules</span> »</span> forbid questioning the status quo. That is, imagine if the
2674 rich seceded from Facebook, and then, instead of running ads that made money
2675 for shareholders, Facebook became a billionaire’s vanity project that also
2676 happened to ensure that nobody talked about whether it was fair that only
2677 billionaires could afford to hang out in the rarified corners of the
2678 internet.
2679   </p><p>
2680     Behind the idea of paying for access is a belief that free markets will
2681 address Big Tech’s dysfunction. After all, to the extent that people have a
2682 view of surveillance at all, it is generally an unfavorable one, and the
2683 longer and more thoroughly one is surveilled, the less one tends to like
2684 it. Same goes for lock-in: If HP’s ink or Apple’s App Store were really
2685 obviously fantastic, they wouldn’t need technical measures to prevent users
2686 from choosing a rival’s product. The only reason these technical
2687 countermeasures exist is that the companies don’t believe their customers
2688 would <span class="emphasis"><em>voluntarily</em></span> submit to their terms, and they want
2689 to deprive them of the choice to take their business elsewhere.
2690   </p><p>
2691     Advocates for markets laud their ability to aggregate the diffused knowledge
2692 of buyers and sellers across a whole society through demand signals, price
2693 signals, and so on. The argument for surveillance capitalism being a
2694 <span class="quote">« <span class="quote">rogue capitalism</span> »</span> is that machine-learning-driven persuasion
2695 techniques distort decision-making by consumers, leading to incorrect
2696 signals — consumers don’t buy what they prefer, they buy what they’re
2697 tricked into preferring. It follows that the monopolistic practices of
2698 lock-in, which do far more to constrain consumers’ free choices, are even
2699 more of a <span class="quote">« <span class="quote">rogue capitalism.</span> »</span>
2700   </p><p>
2701     The profitability of any business is constrained by the possibility that its
2702 customers will take their business elsewhere. Both surveillance and lock-in
2703 are anti-features that no customer wants.  But monopolies can capture their
2704 regulators, crush their competitors, insert themselves into their customers’
2705 lives, and corral people into <span class="quote">« <span class="quote">choosing</span> »</span> their services
2706 regardless of whether they want them — it’s fine to be terrible when there
2707 is no alternative.
2708   </p><p>
2709     Ultimately, surveillance and lock-in are both simply business strategies
2710 that monopolists can choose. Surveillance companies like Google are
2711 perfectly capable of deploying lock-in technologies — just look at the
2712 onerous Android licensing terms that require device-makers to bundle in
2713 Google’s suite of applications. And lock-in companies like Apple are
2714 perfectly capable of subjecting their users to surveillance if it means
2715 keeping the Chinese government happy and preserving ongoing access to
2716 Chinese markets.  Monopolies may be made up of good, ethical people, but as
2717 institutions, they are not your friend — they will do whatever they can get
2718 away with to maximize their profits, and the more monopolistic they are, the
2719 more they <span class="emphasis"><em>can</em></span> get away with.
2720   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="an-ecology-moment-for-trustbusting"></a>An <span class="quote">« <span class="quote">ecology</span> »</span> moment for trustbusting</h2></div></div></div><p>
2721     If we’re going to break Big Tech’s death grip on our digital lives, we’re
2722 going to have to fight monopolies. That may sound pretty mundane and
2723 old-fashioned, something out of the New Deal era, while ending the use of
2724 automated behavioral modification feels like the plotline of a really cool
2725 cyberpunk novel.
2726   </p><p>
2727     Meanwhile, breaking up monopolies is something we seem to have forgotten how
2728 to do. There is a bipartisan, trans-Atlantic consensus that breaking up
2729 companies is a fool’s errand at best — liable to mire your federal
2730 prosecutors in decades of litigation — and counterproductive at worst,
2731 eroding the <span class="quote">« <span class="quote">consumer benefits</span> »</span> of large companies with massive
2732 efficiencies of scale.
2733   </p><p>
2734     But trustbusters once strode the nation, brandishing law books, terrorizing
2735 robber barons, and shattering the illusion of monopolies’ all-powerful grip
2736 on our society. The trustbusting era could not begin until we found the
2737 political will — until the people convinced politicians they’d have their
2738 backs when they went up against the richest, most powerful men in the world.
2739   </p><p>
2740     Could we find that political will again?
2741   </p><p>
2742     Copyright scholar James Boyle has described how the term
2743 <span class="quote">« <span class="quote">ecology</span> »</span> marked a turning point in environmental
2744 activism. Prior to the adoption of this term, people who wanted to preserve
2745 whale populations didn’t necessarily see themselves as fighting the same
2746 battle as people who wanted to protect the ozone layer or fight freshwater
2747 pollution or beat back smog or acid rain.
2748   </p><p>
2749     But the term <span class="quote">« <span class="quote">ecology</span> »</span> welded these disparate causes together
2750 into a single movement, and the members of this movement found solidarity
2751 with one another. The people who cared about smog signed petitions
2752 circulated by the people who wanted to end whaling, and the anti-whalers
2753 marched alongside the people demanding action on acid rain. This uniting
2754 behind a common cause completely changed the dynamics of environmentalism,
2755 setting the stage for today’s climate activism and the sense that preserving
2756 the habitability of the planet Earth is a shared duty among all people.
2757   </p><p>
2758     I believe we are on the verge of a new <span class="quote">« <span class="quote">ecology</span> »</span> moment
2759 dedicated to combating monopolies. After all, tech isn’t the only
2760 concentrated industry nor is it even the <span class="emphasis"><em>most</em></span>
2761 concentrated of industries.
2762   </p><p>
2763     You can find partisans for trustbusting in every sector of the
2764 economy. Everywhere you look, you can find people who’ve been wronged by
2765 monopolists who’ve trashed their finances, their health, their privacy,
2766 their educations, and the lives of people they love.  Those people have the
2767 same cause as the people who want to break up Big Tech and the same
2768 enemies. When most of the world’s wealth is in the hands of a very few, it
2769 follows that nearly every large company will have overlapping shareholders.
2770   </p><p>
2771     That’s the good news: With a little bit of work and a little bit of
2772 coalition building, we have more than enough political will to break up Big
2773 Tech and every other concentrated industry besides. First we take Facebook,
2774 then we take AT&amp;T/WarnerMedia.
2775   </p><p>
2776     But here’s the bad news: Much of what we’re doing to tame Big Tech
2777 <span class="emphasis"><em>instead</em></span> of breaking up the big companies also
2778 forecloses on the possibility of breaking them up later.
2779   </p><p>
2780     Big Tech’s concentration currently means that their inaction on harassment,
2781 for example, leaves users with an impossible choice: absent themselves from
2782 public discourse by, say, quitting Twitter or endure vile, constant
2783 abuse. Big Tech’s over-collection and over-retention of data results in
2784 horrific identity theft. And their inaction on extremist recruitment means
2785 that white supremacists who livestream their shooting rampages can reach an
2786 audience of billions. The combination of tech concentration and media
2787 concentration means that artists’ incomes are falling even as the revenue
2788 generated by their creations are increasing.
2789   </p><p>
2790     Yet governments confronting all of these problems all inevitably converge on
2791 the same solution: deputize the Big Tech giants to police their users and
2792 render them liable for their users’ bad actions. The drive to force Big Tech
2793 to use automated filters to block everything from copyright infringement to
2794 sex-trafficking to violent extremism means that tech companies will have to
2795 allocate hundreds of millions to run these compliance systems.
2796   </p><p>
2797     These rules — the EU’s new Directive on Copyright, Australia’s new terror
2798 regulation, America’s FOSTA/SESTA sex-trafficking law and more — are not
2799 just death warrants for small, upstart competitors that might challenge Big
2800 Tech’s dominance but who lack the deep pockets of established incumbents to
2801 pay for all these automated systems. Worse still, these rules put a floor
2802 under how small we can hope to make Big Tech.
2803   </p><p>
2804     That’s because any move to break up Big Tech and cut it down to size will
2805 have to cope with the hard limit of not making these companies so small that
2806 they can no longer afford to perform these duties — and it’s
2807 <span class="emphasis"><em>expensive</em></span> to invest in those automated filters and
2808 outsource content moderation. It’s already going to be hard to unwind these
2809 deeply concentrated, chimeric behemoths that have been welded together in
2810 the pursuit of monopoly profits. Doing so while simultaneously finding some
2811 way to fill the regulatory void that will be left behind if these
2812 self-policing rulers were forced to suddenly abdicate will be much, much
2813 harder.
2814   </p><p>
2815     Allowing the platforms to grow to their present size has given them a
2816 dominance that is nearly insurmountable — deputizing them with public duties
2817 to redress the pathologies created by their size makes it virtually
2818 impossible to reduce that size. Lather, rinse, repeat: If the platforms
2819 don’t get smaller, they will get larger, and as they get larger, they will
2820 create more problems, which will give rise to more public duties for the
2821 companies, which will make them bigger still.
2822   </p><p>
2823     We can work to fix the internet by breaking up Big Tech and depriving them
2824 of monopoly profits, or we can work to fix Big Tech by making them spend
2825 their monopoly profits on governance. But we can’t do both. We have to
2826 choose between a vibrant, open internet or a dominated, monopolized internet
2827 commanded by Big Tech giants that we struggle with constantly to get them to
2828 behave themselves.
2829   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="make-big-tech-small-again"></a>Make Big Tech small again</h2></div></div></div><p>
2830     Trustbusting is hard. Breaking big companies into smaller ones is expensive
2831 and time-consuming. So time-consuming that by the time you’re done, the
2832 world has often moved on and rendered years of litigation irrelevant. From
2833 1969 to 1982, the U.S. government pursued an antitrust case against IBM over
2834 its dominance of mainframe computing — but the case collapsed in 1982
2835 because mainframes were being speedily replaced by PCs.
2836   </p><div class="blockquote"><blockquote class="blockquote"><p>
2837       A future U.S. president could simply direct their attorney general to
2838 enforce the law as it was written.
2839     </p></blockquote></div><p>
2840     It’s far easier to prevent concentration than to fix it, and reinstating the
2841 traditional contours of U.S. antitrust enforcement will, at the very least,
2842 prevent further concentration. That means bans on mergers between large
2843 companies, on big companies acquiring nascent competitors, and on platform
2844 companies competing directly with the companies that rely on the platforms.
2845   </p><p>
2846     These powers are all in the plain language of U.S. antitrust laws, so in
2847 theory, a future U.S. president could simply direct their attorney general
2848 to enforce the law as it was written. But after decades of judicial
2849 <span class="quote">« <span class="quote">education</span> »</span> in the benefits of monopolies, after multiple
2850 administrations that have packed the federal courts with lifetime-appointed
2851 monopoly cheerleaders, it’s not clear that mere administrative action would
2852 do the trick.
2853   </p><p>
2854     If the courts frustrate the Justice Department and the president, the next
2855 stop would be Congress, which could eliminate any doubt about how antitrust
2856 law should be enforced in the U.S. by passing new laws that boil down to
2857 saying, <span class="quote">« <span class="quote">Knock it off. We all know what the Sherman Act says. Robert
2858 Bork was a deranged fantasist. For avoidance of doubt, <span class="emphasis"><em>fuck that
2859 guy</em></span>.</span> »</span> In other words, the problem with monopolies is
2860 <span class="emphasis"><em>monopolism</em></span> — the concentration of power into too few
2861 hands, which erodes our right to self-determination. If there is a monopoly,
2862 the law wants it gone, period. Sure, get rid of monopolies that create
2863 <span class="quote">« <span class="quote">consumer harm</span> »</span> in the form of higher prices, but also,
2864 <span class="emphasis"><em>get rid of other monopolies, too</em></span>.
2865   </p><p>
2866     But this only prevents things from getting worse. To help them get better,
2867 we will have to build coalitions with other activists in the anti-monopoly
2868 ecology movement — a pluralism movement or a self-determination movement —
2869 and target existing monopolies in every industry for breakup and structural
2870 separation rules that prevent, for example, the giant eyewear monopolist
2871 Luxottica from dominating both the sale and the manufacture of spectacles.
2872   </p><p>
2873     In an important sense, it doesn’t matter which industry the breakups begin
2874 in. Once they start, shareholders in <span class="emphasis"><em>every</em></span> industry
2875 will start to eye their investments in monopolists skeptically. As
2876 trustbusters ride into town and start making lives miserable for
2877 monopolists, the debate around every corporate boardroom’s table will
2878 shift. People within corporations who’ve always felt uneasy about monopolism
2879 will gain a powerful new argument to fend off their evil rivals in the
2880 corporate hierarchy: <span class="quote">« <span class="quote">If we do it my way, we make less money; if we do
2881 it your way, a judge will fine us billions and expose us to ridicule and
2882 public disapprobation. So even though I get that it would be really cool to
2883 do that merger, lock out that competitor, or buy that little company and
2884 kill it before it can threaten it, we really shouldn’t — not if we don’t
2885 want to get tied to the DOJ’s bumper and get dragged up and down Trustbuster
2886 Road for the next 10 years.</span> »</span>
2887   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="goto-10"></a>20 GOTO 10</h2></div></div></div><p>
2888     Fixing Big Tech will require a lot of iteration. As cyber lawyer Lawrence
2889 Lessig wrote in his 1999 book, <span class="emphasis"><em>Code and Other Laws of
2890 Cyberspace</em></span>, our lives are regulated by four forces: law (what’s
2891 legal), code (what’s technologically possible), norms (what’s socially
2892 acceptable), and markets (what’s profitable).
2893   </p><p>
2894     If you could wave a wand and get Congress to pass a law that re-fanged the
2895 Sherman Act tomorrow, you could use the impending breakups to convince
2896 venture capitalists to fund competitors to Facebook, Google, Twitter, and
2897 Apple that would be waiting in the wings after they were cut down to size.
2898   </p><p>
2899     But getting Congress to act will require a massive normative shift, a mass
2900 movement of people who care about monopolies — and pulling them apart.
2901   </p><p>
2902     Getting people to care about monopolies will take technological
2903 interventions that help them to see what a world free from Big Tech might
2904 look like. Imagine if someone could make a beloved (but unauthorized)
2905 third-party Facebook or Twitter client that dampens the anxiety-producing
2906 algorithmic drumbeat and still lets you talk to your friends without being
2907 spied upon — something that made social media more sociable and less
2908 toxic. Now imagine that it gets shut down in a brutal legal battle. It’s
2909 always easier to convince people that something must be done to save a thing
2910 they love than it is to excite them about something that doesn’t even exist
2911 yet.
2912   </p><p>
2913     Neither tech nor law nor code nor markets are sufficient to reform Big
2914 Tech. But a profitable competitor to Big Tech could bankroll a legislative
2915 push; legal reform can embolden a toolsmith to make a better tool; the tool
2916 can create customers for a potential business who value the benefits of the
2917 internet but want them delivered without Big Tech; and that business can get
2918 funded and divert some of its profits to legal reform. 20 GOTO 10 (or
2919 lather, rinse, repeat). Do it again, but this time, get farther! After all,
2920 this time you’re starting with weaker Big Tech adversaries, a constituency
2921 that understands things can be better, Big Tech rivals who’ll help ensure
2922 their own future by bankrolling reform, and code that other programmers can
2923 build on to weaken Big Tech even further.
2924   </p><p>
2925     The surveillance capitalism hypothesis — that Big Tech’s products really
2926 work as well as they say they do and that’s why everything is so screwed up
2927 — is way too easy on surveillance and even easier on capitalism. Companies
2928 spy because they believe their own BS, and companies spy because governments
2929 let them, and companies spy because any advantage from spying is so
2930 short-lived and minor that they have to do more and more of it just to stay
2931 in place.
2932   </p><p>
2933     As to why things are so screwed up? Capitalism. Specifically, the monopolism
2934 that creates inequality and the inequality that creates monopolism. It’s a
2935 form of capitalism that rewards sociopaths who destroy the real economy to
2936 inflate the bottom line, and they get away with it for the same reason
2937 companies get away with spying: because our governments are in thrall to
2938 both the ideology that says monopolies are actually just fine and in thrall
2939 to the ideology that says that in a monopolistic world, you’d better not
2940 piss off the monopolists.
2941   </p><p>
2942     Surveillance doesn’t make capitalism rogue. Capitalism’s unchecked rule
2943 begets surveillance. Surveillance isn’t bad because it lets people
2944 manipulate us. It’s bad because it crushes our ability to be our authentic
2945 selves — and because it lets the rich and powerful figure out who might be
2946 thinking of building guillotines and what dirt they can use to discredit
2947 those embryonic guillotine-builders before they can even get to the
2948 lumberyard.
2949   </p></div><div class="sect1"><div class="titlepage"><div><div><h2 class="title" style="clear: both"><a name="up-and-through"></a>Up and through</h2></div></div></div><p>
2950     With all the problems of Big Tech, it’s tempting to imagine solving the
2951 problem by returning to a world without tech at all. Resist that temptation.
2952   </p><p>
2953     The only way out of our Big Tech problem is up and through. If our future is
2954 not reliant upon high tech, it will be because civilization has fallen. Big
2955 Tech wired together a planetary, species-wide nervous system that, with the
2956 proper reforms and course corrections, is capable of seeing us through the
2957 existential challenge of our species and planet. Now it’s up to us to seize
2958 the means of computation, putting that electronic nervous system under
2959 democratic, accountable control.
2960   </p><p>
2961     I am, secretly, despite what I have said earlier, a tech exceptionalist. Not
2962 in the sense of thinking that tech should be given a free pass to monopolize
2963 because it has <span class="quote">« <span class="quote">economies of scale</span> »</span> or some other nebulous
2964 feature. I’m a tech exceptionalist because I believe that getting tech right
2965 matters and that getting it wrong will be an unmitigated catastrophe — and
2966 doing it right can give us the power to work together to save our
2967 civilization, our species, and our planet.
2968   </p></div></div></body></html>