Tuesday, May 22, 2012
Are there practical remedies that wouldn’t involve federal regulation of search results? Comments of Frank Pasquale
Posted by Frank Pasquale (Seton Hall)
Could Google make a mistake? Could the company itself, or a small group of its thousands of employees, act with motives that diverge from its publicly stated mission? If such actions hurt the owners of a website, should they in some cases be able to seek recourse from some other entity than Google itself? I believe the answer to all these questions is yes. Answering them helps us confront issues of discrimination, malfeasance, nonfeasance, and technological due process in a rapidly changing online environment. They also suggest how Google might better respond to ongoing investigations and concern about its practices.
Imagine that you own Company A, and your main competitor is the persistent (but demonstrably worse) Company B. In searches for the products you sell, you reliably end up in the top five results in the studies you’ve commissioned; your competitors at Company B are on the fifth or sixth pages.* What happens if Google purchases Company B, and immediately after the purchase, Company B appears to dominate the first page of results, and your company has been relegated to later pages? Stipulate that repeated appeals to webmaster forums and other mechanisms of corporate due process fail. Should there be some type of remedy?
This may seem like a crude and overdrawn hypothetical, but it is a good heuristic for testing the extraordinary skepticism toward independent evaluation of Google's performance and motives expressed in the Ammori/Pelican paper prepared for Google. (For those who prefer more subtle scenarios, please consult my discussion of Google's potential bias against rival video sites after it purchased YouTube, in the article Internet Nondiscrimination Principles. The recently released Statement of EC VP Almunia on the Google antitrust investigation also evokes relevant scenarios.) After reviewing Ammori & Pelican’s work, I worry they would oppose any remedy at law for the owner of Company A. I was particularly troubled by their critique of third-party investigative committees (be they government agencies or nongovernment organizations) which would be able to understand how decisions to change ranking methods originated and how they were implemented.
Qualified Transparency as Remedy
Changes in ranking methodology are rigorously tested and documented. When a website suddenly tumbles dozens of places, and has a plausible story about being targeted as a potential rival of an established Google interest or a space the company is planning to invest in, is it too much to ask for some third party to review the particular factors that led to the demotion? Given how quickly a sudden drop can occur, we are not discussing an infinite variety of changes to be reviewed. Nor are we demanding the disclosure of the entire algorithm to a third-party auditor, or even the revelation of the relevant changes in the algorithm to the party involved, much less the general public. In my early work on this topic, my co-author and I even pointed to the precedent of the supersecretive FISA court as a model, to underscore how much we respected the intellectual property rights of the company (and particularly the value of trade secrecy).
As Mark Patterson has shown, the Dodd-Frank Act already requires far more disclosure from rating agencies. Patterson makes these key points:
Google has been alleged to have manipulated its search results (or ratings) in much the same way that the rating agencies have been alleged to have manipulated credit ratings. Although Google is alleged to have manipulated the ratings of competitors (e.g., potentially competing “vertical” search engines) and credit rating agencies are alleged to have manipulated the ratings of customers (issuers of financial products), the basic phenomenon is the same. . . . lack of transparency in quality can give an information provider market power, as does an absence of price transparency.
[A] requirement imposed on credit-rating agencies in the recent Dodd-Frank financial reform legislation is also well-suited to address competition issues [involving information providers]. In Dodd-Frank, Congress directed the SEC to prescribe rules that, when credit-rating agencies make ‘material changes’ to ‘rating procedures and methodologies,’ ensure that: ‘the changes are applied consistently to all credit ratings to which the changed procedures and methodologies apply. . . and [the CRA] publicly discloses the reason for the change . . .’
Such requirements do not impinge on the information providers’ editorial judgment; they simply require that some information about it be given.
I have discussed other potential models for detecting anti-competitive biases. In contexts ranging from privacy rights to false advertising, authorities in the US and Europe have recognized the need for fast, flexible “quick looks” at suspect business practices. In one of the scenarios I have mentioned involving the FTC and the NAD, 95% of problematic situations are quickly resolved in a self-regulatory fashion. This is not a recipe for the litigation nightmares industry advocates so frequently invoke.
The institutions devoted to fair data practices in Europe that I mention in the Northwestern piece may at first seem inapposite here; complainants like Foundem want more, not less, exposure. However, at bottom the two types of disputes share a critical element: each aggrieved party may feel that the correct data about it has not been logged, used, or processed accurately. We now need to decide whether some entity outside of Google, whether as a result of competition law complaints, consumer protection law, or other principles of commercial fairness, has authority to review and offer its judgment on such questions.
What that entity ultimately does about its findings is not my central concern at this time. Simply informing consumers about potential biases would be a valuable public service. Search is often a credence service, and in areas ranging from health care to law we recognize the need for third parties to verify and validate the quality of credence services provided. In past work, I have proposed a limited right of annotation (as minor as an asterisk) for certain entities aggrieved by problematic search engine rankings.
Both disclosure and annotation remedies can be implemented in many different ways, with whatever depth and breadth the situation may warrant. As Ayres and Braithwaite demonstrated two decades ago, “government can support and encourage industry self-regulation.” Is it too much to ask for some entity outside of Google to be able to “look under the hood” and understand what is going on in plausibly contested scenarios? If so, such an abdication of administrative responsibility in the face of technical complexity bodes ill not merely for a level competitive playing field, but for democratic and judicial processes themselves. Antitrust law flirts with irrelevance if it disdains the technical tools necessary to understand a modern information economy.
Avoiding Digital Feudalism
Disclosure and auditing are not merely remedies; they may even be considered salutary extensions of current business practices. Search engines rank websites at least in part by algorithmically processing signals from the site (such as the number of links to it, the number of links to those links, whether users have clicked on the site when it was ranked in search results previously, etc.). Let's say that, for whatever reason, Google's technology fails to pick up on all the signals relating to a given website. The webmaster for the site might complain, and may well get a response. Establishing webmaster forums that allow for that type of dialogue between the ranked and the ranker has, to this point at least, seemed like a fair and responsible business practice to Google itself.
If Google thought of its rankings as a kind of virtual world, whose members have essentially accepted (via terms of service) the absolute sovereignty of the ruler of that territory in the metaverse, such a dialogical process would make little sense. In the digital feudalism of virtual worlds, no one has a right to question the unilateral decision of the ruler. Errors would only have meaning as lost profit opportunities, not as failures to run a competition properly. At its best, Google recognizes itself in the latter role, rather than as a purely profit-maximizing entity.
Google is not alone in exercising power over the internet. Apple, Facebook, Twitter, and Amazon can also rely on opaque technologies, sometimes leaving users in the dark as to exactly why any given app, story, or book is featured at a particular time. Though we are focusing on Google now, the problems raised in its various antitrust disputes are not isolated. Rather, we should expect any company aspiring to order vast amounts of information to try to keep its methods secret, if only to reduce controversy and foil copycat competitors. However wise this secrecy may be as a business strategy, it devastates our ability to truly understand the social world Silicon Valley is creating. Moreover, like a modern-day Ring of Gyges, opacity creates ample opportunities to hide anti-competitive, discriminatory, or simply careless conduct behind a veil of technical inscrutability. Qualified transparency can address these concerns while respecting intellectual property rights.
It is disappointing to see Marvin Ammori, who has done so much to bring carriers’ troubling practices to light (and to justice), ignore parallel problems in other parts of the internet. If policymakers accept the arguments he is making now with respect to Google, he may undermine all he has accomplished in the realm of net neutrality. Bottlenecks at any layer of the Internet create opportunities for the exercise of undue power over the flow of information and ideas.
* I don’t even raise the possibility of the site owner knowing whether it is in the top 5 results generally, because, in an era of personalization, only Google can know that. At present, we can only hope for relatively good sampling of sites.