If you’d like an essay-formatted version of this post to read or share, here’s a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:
Last August, a federal judge convicted Google of being “a monopolist” and acting “as one to maintain its monopoly.” The judge concluded that key to Google’s monopoly was the vast troves of data it collects and analyzes and asked the parties to come up with remedies to address this.
Many trustbusters and Google competitors read this and concluded that Google should be forced to share its click and quer y data. The technical term for this is “apocalyptically stupid.” Releasing Google’s click and query data into the wild is a privacy Chernobyl in the waiting. The secrets that we whisper to search engines have the power to destroy us a thousand times over.
Largely theoretical answers like “differential privacy” are promising, but remain theoretical at scale. The first large-scale live-fire exercise for these should not be something as high-stakes as Google’s click and query data. If anything, we should
delete that data:
The last thing we want to do is use antitrust to democratize surveillance so that everyone can spy as efficiently as Google does. In theory, we could sanitize the click and query data by limiting sharing to queries that were made by multiple, independent users (say, only sharing queries that at least 30 users have made), but it’s unlikely that this will do much to improve the performance of rival firms’ search engines.
Google only retains 18 months’ worth of click and query data, thus once we cut off its capacity to collect more data, whatever advantage it has from surveillance will begin to decay immediately and fall to zero in 18 months.
(However: the 18 months figure is deceptive, and deliberately so. Google may only retain your queries for 18 months, but it is silent on how long it retains the
inferences from those queries. It may discard your “how do I get an abortion in my red state” query after a year and a half, but indefinitely retain the “sought an illegal abortion” label it added to your profile. The US
desperately needs a federal consumer privacy law!)
And just to be clear, there’s
other Google data that would be very useful to rival search engines, like Google’s search index – the trove of pages from the internet. Google already licenses this out, and search engines like Kagi use it to produce
substantially superior search results:
The DOJ has just filed its proposal for a remedy, and it’s a doozy: forcing Google to sell off Chrome, on the basis that both of these are the source of much of Google’s data, and no rival search engine is likely to also have a widely used browser:
This represents something of a compromise position: the DOJ had initially signalled that it would also demand a selloff of Android, and that’s been dropped. I think there’s a good case for forcing the sale of Android as a source of data, too.
In competition theory, these selloffs are referred to as “structural separation” – when a company that provides infrastructure to other firms is prohibited from competing with those firms:
For example, it used to be that banks were prohibited from competing with the companies they loaned money to. After all, if you borrow money from Chase to open a pizzeria, and then Chase opens a pizzeria of its own across the street, you can see how your business would be doomed. You have to make interest payments to Chase, and your rival doesn’t, and if Chase wants to, it can subsidize that rival so it can sell pizzas below cost until you’re out of business.
Likewise, rail companies were banned from owning freight companies, because otherwise they would destroy the businesses of every freight company that shipped on the railroad.
In theory, you could create fair play rules that required the bank or the railroad to play nice with the business customers that used their platforms, but in practice, there are so many ways of cheating that this would be unenforceable.
This principle is well established in all other areas of business, and we recoil in horror when it is violated. You wouldn’t hire a lawyer who was also representing the person who’s suing you. Judges (with the abominable exception of Supreme Court justices!) are required to recuse themselves when they have a personal connection with either of the parties in a case they preside over.
One of the weirdest sights of the new Gilded Age is when lawyers for monopoly companies argue that they can play fair with their customers despite their conflicts of interest. Think of Google or Meta, with their ad-tech duopoly. These are companies that purport to represent sellers of ads
and buyers of ads in marketplaces they own and control, and where they compete with sellers and/or buyers. These companies suck up 51% of the revenue generated by advertising, while historically, the share taken by ad intermediaries was more like 15%!
Imagine if you and your partner discovered that the same lawyer was representing
both of you in the divorce, while also serving as the judge,
and trying to match with both of you on Tinder. Now imagine that when the divorce terms were finalized, lawyer got your family home.
No Google lawyer would agree to argue on the company’s behalf in a case where the judge was employed by the party that’s suing them, but they will blithely argue that the reason they’re getting 51% of the ad-rake is that they’re providing 51% of the value.
Structural separation – like judicial recusal – comprehensively and unarguably resolves all the perceptions and realities of conflict between parties. The fact that platform owners compete with platform users is the source bottomless corruption, from Google to Amazon:
ALT
