Communications of the ACM

Law and technology

What to Do About Google?

By James Grimmelmann
Communications of the ACM, September 2013, Vol. 56 No. 9, Pages 28-30
10.1145/2500129
Comments

Google is a noun, a verb, and a controversy. It receives two-thirds of all searches in the U.S., and more than 90% in many European countries. It has dipped its toes—or perhaps its tentacles—into local listings, news, books, videos, flights, patents, and prices, to name just a few. If it exists, Google wants to index it.

Unsurprisingly, this modern octopus has its critics. There are, among others, newspapers upset about having their headlines scraped and aggregated, trademark owners upset about keyword ads for their competitors, introverts upset that searches on their names resurrect painful and humiliating memories, governments upset at the subversive and scandalous things citizens can find with a quick search, and privacy advocates upset at Google's immense stockpiles of personal data. And that is just the search engine; if one were to add in the concerns about Android rootkits, Google Glass creep shots, driverless car crashes and the rest, this column would not be long enough to list them all.)

The oldest and most persistent critique of Google's power, known as "search bias," is the fear that search rankings create reality rather than reflecting it. If Google demotes the restaurant Le Snoot from being the first result for "restaurant near 54321" to the hundreth, many gourmets will make their reservations elsewhere. If Dave's Diner takes its place as the number-one result, diners will go there instead. Google can literally pick winners and losers in the game of the Internet.

The most explosive search bias allegations against Google involve its vertical search engines, like Google+Local (to find nearby businesses and restaurants) and Google Flights (to find and book airline tickets). Google gives these specialized search results prominent placement at the top of its results pages. Competitors like Yelp and Expedia have charged that this gives Google's vertical offerings an unfair advantage over their competing vertical search engines. For years, they and other Google critics have been pressing regulators in the U.S. and the European Union to curb Google's allegedly abusive practices.

---

Google can literally pick winners and losers in the game of the internet.

---

But they have been sorely disappointed, on both sides of the Atlantic. In January, the Federal Trade Commission dropped its search bias investigation, concluding the changes competitors complained about "could plausibly be viewed as an improvement in the overall quality of Google's search results."⁵ In April, the European Commission went a bit further, but not much. In a proposed settlement, Google agreed to label its own vertical search results more prominently, and to add a few, not particularly conspicuous, links to rival vertical search en- gines.⁶ Both regulators left untouched the core practice responsible for so much criticism: top-ranked placement for Google's own news, flight information, and local results.

Did the authorities shirk their responsibilities to rein in an unruly titan? Or did they show admirable restraint in refusing the gum up the gears of an innovative technology? It is impossible to answer these and other policy questions about Google without some theory of what search engines are good for and what society ought to expect of them.

Fortunately, we have such a theory—or rather, we have three such theories. Some observers have compared Google to a traditional telecommunications conduit like a radio station.² Some have compared it to an editor deciding what stories to put in a magazine.⁸ And some have compared it to an advisor, like the concierge in a hotel who answers questions about local attractions.⁷ Each theory offers its own insights.

Calling search engines conduits emphasizes that they have become one of the new bottlenecks on the Internet. If Le Snoot's ISP decides to unplug its connection, no one will be able to reach lesnoot.com. The same will be true if the DNS records for lesnoot.com are deleted, or if search engines drop lesnoot.com from their indexes. And so, if the parallel holds, just as the Bell telephone network was regulated to ensure nondiscriminatory access for everyone, search engines should be too.

When people talk about "search neutrality"—by analogy to "network neutrality"—they are making an argument for treating Google as a conduit. Of course, Google could not simply rank all websites identically, because only one result can be first, but it ought to treat them all fairly. The opposite of a "neutral" search engine is a "biased" search engine; rather than listing websites in the order they deserve to be ranked, it injects its own discriminatory distortions. The claim that Google is doing something wrong when it puts its own flight search results higher on the page than Expedia's is a claim that Google should be acting as a neutral conduit but is not.

The conduit theory's natural enemy is the editor theory, which says that making distinctions among web-pages is an act of judicious discretion rather than dangerous discrimination. The editors of Communications make countless decisions about which developments in computing are worth covering, which articles are most informative, and where to put them in the magazine. These decisions are not "right" or "wrong"; they simply reflect the judgment of its editorial board and staff. Google sees itself the same way. True, Google's editorial cycle is measured in milliseconds rather than in months. But when its search quality team meets to discuss algorithmic tweaks, it resembles a newspaper staff debating which stories to put on the front page of the metro section. And, continues the argument, just like the government cannot tell the New York Times to spike an unflattering story about Guantanamo Bay, it cannot tell Google which search results to show.

Finally, one could view Google as an advisor, helping users find what they are looking for. If so, the best search engine is one that is most useful to users, rather than the one that is least biased, or most reflects its programmer's point of view. Le Snoot may be the "best" restaurant in town, as judged by professional food critics. But some people do not like heavy French cuisine, others are vegans, and even cassoulet lovers would rather just have a slice of pizza now and then. Whether Le Snoot or Dave's Diner is more relevant to a user depends on what she intends as she types her query.

Calling Google an advisor cuts both ways: it gives Google both rights and duties. It gives a powerful argument against search neutrality: a law that puts Le Snoot back on top makes it more difficult for the user who wants a grilled cheese sandwich to get a decent meal. But just as readers would rightly be furious to discover the hotel concierge only recommended Le Snoot because the head chef slipped him an envelope stuffed with cash, search users would also have cause to complain if payola determined search rankings. More than a decade ago, the FTC strongly warned search engines against displaying undisclosed paid listings.⁴

All three theories capture something important about how search engines work. Each of them celebrates the contributions of one of the essential parties to a search. The conduit theory is all about websites with something to say, the advisor theory is all about the users who are interested in listening, and the editor theory is all about the search engine that connects them.

But when it comes to crafting sensible law for search engines, our sympathies should lie with users. The Internet has made it easier to speak to worldwide audiences than ever before, but at the cost of massively increasing the cacophony confronting those audiences. Since users' interests are as diverse as human thought, they need highly personalized help in picking through the treasures in the Internet's vast but utterly disorganized storehouse. The search engine is the only technology known to humanity capable of solving this problem at Internet scale.

Some familiar controversies about Google look rather different from this point of view. Take search bias. If Google is a conduit, bias is a serious problem; Google is setting up orange cones to block the highway and divert Internet users to the Google exit. If Google is an editor, bias is just as much a non-issue as when the front page of the Daily News promotes its own sports coverage rather than the Post's.

---

When it comes to crafting sensible law for search engines, our sympathies should lie with users.

---

If Google is an advisor, though, the answer lies somewhere between "always wrong" and "always fine." The key question is not whether Google is helping itself or whether it is hurting websites, but whether it is helping users find what they want. Sometimes, for some queries, Google can quite reasonably think that users will be grateful if it lists its own services first. Flight search is a good example: Google's interactive OneBox helps users dive right into the flight-picking process.

At other times, for other queries, Google may have strong evidence that users prefer particular sites. If Google demotes them to insert its own pages that it knows users would rather not see, that could be problematic. It is a form of deception: Google is telling the user, "This is the best I can do for you" when it knows full well it could do better.

The FTC properly recognized that deception was the real issue in the Google case. The FTC's decision to drop its search bias investigation hinged on a conclusion that Google had not underplayed its hand. Some, like Expedia and Yelp, criticized the outcome. But there is a difference between disagreeing with Google's ranking decisions—everyone wants to be king of the results page—and showing that those decisions were made in bad faith.

Another advantage of treating search engines as advisors is that it helps put user privacy at the center of the conversation. We depend on advisors to keep confidential what we tell them: doctors and investment advisers are legally obligated to secure their records; so too for search engines. Our query histories are some of the most personal and potentially embarrassing data trails we leave behind us. They have even been used as evidence in murder trials. Strong privacy protections for user search data are essential.

Some of these points apply beyond search engines; some do not. The anti-payola principle is a general one; the FTC has warned advertisers that they must disclose sponsored blog posts, and even sponsored tweets.³ So is the idea that the government should not make users' choices for them; Tulsa cannot tell Yelp that the Holiday Inn deserves an extra star and the Ramada does not. But the duty of loyalty is weaker where advice is not personalized; consumers can continue to leave humorous reviews of the Three Wolf Moon T-shirt at Amazon, even though the reviews may not be especially helpful for shoppers.¹

Google is not the Eye of Sauron, finding all that is good on the Internet and corrupting it. Nor, despite its mission "to organize the world's information and make it universally accessible and useful," is it humanity's informational savior. Google is a company that provides an enormously significant online service. When that service raises serious legal questions, we should ask whether it is good for the users or bad for the users.

Back to Top

Back to Top

The Digital Library is published by the Association for Computing Machinery. Copyright © 2013 ACM, Inc.