Day 9: Perfectionism on Trial
We finally hear from Google's star technical expert, who underscores aggressive assumptions underlying Google's position on divestitures.
Day nine, and the Google ad tech remedy trial is winding down. At end of day, we were halfway through the testimony of Google’s star final witness, Jason Nieh, Professor of Computer Science at Columbia University. He will finish tomorrow, and only rebuttals will remain.
The day started with a Google motion to dismiss the testimony of one of DOJ’s rebuttal witnesses, Stephanie Layser. Denied. Then there was a brief video redirect from George Levitte, who testified on day six, defending Google (and himself) from Rajeev Goel’s claim that Google has left uncorrected a “bug” that prevented a large social media publisher from Open Bidding through Goel’s firm, Pubmatic. Levitte says there were multiple problems but claims they have been working diligently on them, and informing Goel all along.
Then we heard from Heather Adkins, Google’s VP for Security Engineering. But I’m going to focus on one issue in Nieh’s testimony, so I’ll skip over Adkins and the completion of Andres Lerner’s cross-examination from day eight. An excellent and thorough summary of both is at US v Google. Through the lens of security, Adkins revisited the Google defense main theme: Google’s source code is staggeringly immense and inextricably interconnected—illustrated this time by the metaphor of her interlaced fingers—making divestiture of AdX and/or DFP unimaginably difficult. And Wood’s cross of Lerner is very worth reading.
The reason I want to focus on one moment from Nieh (well, two) is that I think he illuminates—maybe even resolves!—the most intractable dispute of this remedy trial: the question of divestiture’s technical feasibility. The important moment is early in Nieh’s testimony, when he summarizes his report. Here are the assumptions he made for his report (numbering added):
1. Divested functionality must have equivalent features and scaling capabilities as GAM’s ad exchange (AdX) or ad server (DFP) functionalities do now.
2. As part of the process of divestiture, any disruption to customers should be minimized.
Jeannie Rhee asked Nieh if these differ from DOJ’s assumptions and how. He replied that DOJ does not assume these at all.
Until this statement by Nieh, I had not heard any witness for either side say explicitly that DOJ does not share these assumptions. I have heard the opposite: in the course of his day eight rebuttal of Robin Lee, Lerner says that Lee assumes AdX can be divested while retaining the same commercial value it has now. I noted at the time that I didn’t think Lee had said that in his testimony; but since Lerner had studied Lee’s report, perhaps he had more information. On the other hand, claiming your opponent agrees with you on some point is a tactic for making that point seem less controversial than it is. Shane Goodwin, too, seemed to be trying to sneak in the assumption that “a successful divestiture” had to maintain asset value. I wrote yesterday that DOJ should not feel constrained by that assumption, and now Nieh has confirmed that they are not.
The real significance of Nieh’s assumption one, and the DOJ’s rejection of it, came out when he summarized his report’s conclusions. Nieh’s first conclusion was that divestiture of either AdX or DFP presented unprecedented complexity, high uncertainty, and no guarantee of success, which Nieh, like Goodwin, defined as preserving existing functionality, or as close as possible. But then he made another important point fully explicit: he said, “Sure, you can provide some functionality, some sort of exchange. But providing the level of functionality of AdX today, that’s the complexity.” Suddenly, it all became clear.
Despite the framing that both sides have used throughout this remedy trial, the question is not “Is divestiture technically feasible?” As Nieh has just said, the answer to that question is easy: of course it’s feasible! The real question is the quality of the final product, and that question is on a sliding scale. Google is fighting furiously to embed Nieh’s assumption one—that full functionality must be maintained—into this remedy trial, so that anything short of a perfect AdX will seem unacceptable to the court.
I started in yesterday’s post to argue that a properly holistic and rugged capitalism doesn’t care about perfection. However good or bad the final product is will be priced in, and whatever level of quality it achieves will hurt some people and benefit others. A divested AdX that is less perfect will help its rivals, and presumably will speed the decline of its still-massive market share, evening out competition.
But why has DOJ seemed to accept that the goal must be “providing the level of functionality of AdX today”? If Nieh is right (as I’m sure he is) that DOJ does not share his assumption one, then they have not officially accepted an obligation to only divest Google if they can do it perfectly. But they’re keeping pretty quiet about their right to settle for less than perfect. Why? I think it is a matter of both strategy and psychology.
The basic idea is that you have to fight every issue your opponent fights, if you don’t know what the judge thinks is important. Strategically, if DOJ does not know Judge Brinkema’s tolerance for degrading AdX in the process of divesting it, they cannot aim below perfection, even if they themselves believe, say, that a divested AdX with 75 percent of today’s functionality would be easier and faster to produce and better for restoring competition to the ad exchange market. Psychologically, no one wants to be seen as giving less than their best effort.
I have characterized the idea that preserving AdX’s commercial value is required for a “successful” divestiture as a partial or one-sided view of capitalism. It considers the economic interests of Google and of the buyer of AdX, but not their exchange market competitors, let alone the lucky scavengers buying up the assets of a failed AdX dirt cheap at auction. This partial view can seem natural—look how readily we all think of the stock market as a measure of our economy’s health. The naturalness of that view. though, is the product of a long historical process of institutionalizing it, generally associated with the “Chicago School” in economics. This a huge topic, but worth mentioning here because we saw it in action in this trial in Lerner’s testimony.
Lerner is a particularly extreme practitioner of this style of economics—or particularly “pure,” if “extreme” sounds pejorative. The extent to which his view has been institutionalized can be seen in his advocating for the benefits of monopoly, arguing that the judge should not even have the goal of eliminating Google’s monopoly, if it would have occurred anyway in his “But-For World.” Merely having a monopoly is not per se illegal if you came by it honestly, but using anticompetitive tactics to create or maintain a monopoly is illegal—and there are consequences for illegal conduct. Lerner verged on relitigating liability, suggesting that a monopoly produced by efficiency should be considered, per se, legal.
Parts of this view, I would argue, are simply in error. I have tried to get this across by saying that a holistic capitalism views the interests of all people equally, including those who benefit from failure. So, the investment banker view, voiced by Goodwin, that preserving asset value is not just in the interest of the buyer and seller of the asset but also of the overall society (the court’s concern) ignores relevant interests. Analysis that makes sense for an individual case—such as the sale of a particular asset, which concerns mainly just the parties to the sale—does not apply to the whole society. The error of thinking is seen in the idea that government should be run like a business. This error was most famously seen in the government’s initial response to the 1929 Great Depression, when it made the mistake of tightening its belt, like a business does in hard times, instead of opening its purse, like a government must in hard times. (As I said, it’s a big topic). In any event, the intent of the antitrust laws clearly tips in favor of trusting the competitive process.
The upshot that is even though he is Google’s witness, Nieh has strengthened the DOJ’s chances of a court-ordered divestiture by so bluntly clarifying that of course it can be done. All DOJ has to show is they can do it well enough to serve the court’s purposes.