Google has frequently been accused of search bias, that is, of presenting its search results in an unfair and subjective order. What does that mean? And why should anyone care?
What is search engine bias?
In one sense bias truly is the essence of any good search engine. This is Google's argument and the argument of many of its defenders. Bias simply means placing some websites ahead of others, which of course is a geometric necessity, since only one website can be first, or second, or third. Google also argues, correctly, that deciding who should be first is the very essence of a good search engine, and the sites that are ranked second, or forty-second, are going to claim that they are victims of bias. This seems to be more accurately seen as a definition of website ranking.
Let's define ranking bias, or simply bias, as follows:
Search Bias is the arbitrary ranking of websites to advance the search engine provider's own commercial interests. Bias can take many forms: (1) providing preferential treatment for a site owned by the search engine provider; (2) providing preferential treatment for a site commercially aligned with the search engine provider as an advertiser or in some other way; (3) lowering the ranking of a site that competes with the search engine provider in any of its businesses; or (4) lowering the ranking of a site that has complained about the search engine provider's business practices, or provides news coverage of an event that the search engine provider would prefer not to have publicized.
What does search engine bias look like?
We have come to accept some forms of bias. We expect paid search to be biased in favor of successful bidders; otherwise why would anyone pay for keywords? We expect paid search not to be too biased in favor of bidders, otherwise why would anyone trust sponsored sites on the top of the search list? Google has balanced the pressure for bias, but not for too much bias, by combining a site's bid and the site's quality score. The bid and quality score yield a combined ranking score, which is called rank by relevance by Google and called rank by revenue by other analysts. A site that has an overall quality score ten times higher than a competing site would presumably need to bid only 10% as much to have an equal combined ranking score. Under rank by relevance, the higher quality site is given a lower price for the top spot, since it is the spot most desired by consumers. Alternatively but equivalently, under rank by revenue the higher quality site once again is given a lower price for the top, because doing so is most profitable for Google. And yet, an inferior site, one that sells counterfeit, defective, or expired products, should have much lower product costs and much higher gross profits per click, and therefore might be able to bid far more for a top spot. When this happens an inferior site becomes more profitable for Google, and is allowed to buy the top.
At some price per click bids from each, Google should be indifferent between the two websites; should we be indifferent as well? The recent Non-Prosecution agreement, the $500 million settlement, and the scandal over Larry Page's involvement all suggest that Google's interests and those of the larger public may not be fully aligned in paid search.
Consumers do not expect deliberate manipulation of search order when reporting the results of organic (free) search, which would make bias of organic search both deceptive and commercially effective. Manipulation of organic search order would include altering rankings through any mechanism, manual or algorithmic, to advance Google's own business interests.
There are numerous factors that would need to be assessed to determine if Google's secret black box algorithms are fair or not, including a determination of what fair would mean, and including a determination of what Google actually does. Ben Edelman of Harvard shows data that he believes demonstrates bias; Danny Sullivan of SearchEnglineLand disagrees. Without seeing inside Google's algorithms it's hard to be sure, although there is currently ongoing litigation in the EU that alleges deliberate manipulation to advance Google's own offerings. Other forms of bias are harder to detect. What if Google advances the ranking of websites that have large numbers of ads on them? What if Google allows websites to increase in organic ranking because they have increased in total hits, which can be strongly influenced by their visibility in paid ranking? This last practice would be a strong incentive to pay for keywords, simply to (indirectly) increase ranking in organic search.
We simply don't know. We have allegations and litigation, but we simply don't know if Google search is unfairly manipulated. Again, why should we care? Wright and Manne are quite explicit -- they are not concerned with manipulation unless one can clearly demonstrate harm to consumers.
When is search engine bias a deceptive practice, harmful to consumers?
The manipulation of paid search has the potential to be harmful to consumers. There is a large body of literature that suggests that prominent advertising is indeed assumed by many consumers to be an indication of quality. A provider that features on the top of Google search may therefore be seen as a high quality provider by many consumers. Surely a high ranking in paid search must be credible to and must influence the behavior of a large number of consumers, since if this were not true then selling keywords would not succeed as Google's principle source of revenue.
Misrepresentation or bias in organic search is likely to be most damaging to consumers, in large measure because consumers have come to expect neutrality in organic search. Indeed, they have been promised such neutrality. This is promised as the first of ten things listed in Google's own statement of its mission and its guiding principles:
1. Focus on the user and all else will follow.
Since the beginning, we've focused on providing the best user experience possible. Whether we're designing a new Internet browser or a new tweak to the look of the homepage, we take great care to ensure that they will ultimately serve you, rather than our own internal goal or bottom line. ... Placement in search results is never sold to anyone ...
Admittedly giving something to yourself in organic search, or demoting a competitor in organic search, does not violate the precise wording of the paragraph above, but it surely violates my reading of it.
Isn't this just normal business practice?
Suppose Google is found to promote its own services ahead of services from others, promoting YouTube videos ahead of those from other video services and demoting competitors like Foundem. Suppose Google is found to punish firms that compete directly with it, or that complain about its business practices. Why is this a problem? Everyone knows that if you want an executive program from the Harvard Business School, the Wharton School, or a conference from the Financial Times you go to their websites; everyone knows that you do not learn about a Wharton program at a Harvard website. Why is this any different? Why should Google be the place to learn about competitors' browsers, video services, or Cloud offerings?
First, in some sense Google comes close to being a common carrier, although it is not yet regulated as one. With near-monopoly power in search, it should have obligations to provide some degree of neutrality. This is strengthened by consumers' expectations of neutrality.
Most importantly, if consumers have been promised fair search and are not getting it, this is deceptive. If this increases consumer costs or decreases consumer choice, or if it increases sellers' costs in a way that is passed on to consumers, then consumer harm results.
Is bias automatically self-correcting?
Is it impossible for bias to exist or to continue when choice is one-click away? For a variety of theoretical reasons we would expect that search bias is almost certainly not self-correcting unless it becomes extreme. Surely Google understands this, and surely Google constantly tests for and monitors consumer acceptance. There is a range of reasons why consumers might not detect or react to significant and potentially harmful search bias. This is addressed more fully by Oren Bracha and Frank Pasquale.
Consumers would need to know that what they were getting was demonstrably inferior, and often they do not. Google search is seldom so blatantly biased as to be obvious. If consumers get something adequate, and they do not know the full range of choices, they may not detect that something better is missing.
There is a constant drumbeat of posts claiming that if Google search were biased then it would be demonstrably inferior, and if it were demonstrably inferior then Google would fail. The historical record in a range of other commercial search sites, from Sabre and Apollo in the 1980s to hotels.com more recently, suggests that this is not true.
Summary -- What should the Subcommittee ask Eric Schmidt?
Clearly, the manipulation of search could be harmful to consumers. Could is a very wimpy word to use before a hearing, but hearings come before investigations, investigations precede any litigation, and litigation precedes any findings of fact or findings of law. The point of the Subcommittee Hearings and the FTC investigation are to determine if there is reason to believe that consumers have been harmed.
I would want to see the following questions addressed in Eric Schmidt's testimony:
- To what extent does Google monitor consumers' understanding of Google's business practices? Is there a reason why the purchased spots on the top of the pages were originally labeled "sponsored links" rather than "ads"? If so, was it because consumers trusted sponsored links more, because at the time they did not know that they had been purchased? Do consumers always know when a spot has been purchased today? Is this true of purchased spots in the middle of the page, or of spots on maps? Will this be true of mobile messages sent to consumers in response to contextual cues like texting friends about dinner?
- Do consumers know that search is not really free to them? Do they understand the value of the information that they are disclosing about themselves to Google, and the compensation that they could receive for that information elsewhere? Do they care?
- Under Google's rank by revenue business model it is indeed possible for inferior sites to achieve high placement if they are willing to pay enough; this occurs even if Google is aware of the sites' low quality score. To what extent is Google aware of the possibility of direct and immediate consumer harm? To what extent does Google take steps to ensure that no truly harmful or truly dangerous entity is allowed to bid its way into a prominent spot?
- To what extent does Google actually preference its own businesses? To what extent might this harm competitors' ability to compete, now or in the future?
- As Google and the rest of the search industry moves beyond 10 blue links and begins to offer their own retailing sites, the prospects for preferencing their own sites becomes greater, and more dangerous. I have already written about the potential danger of Bing's travel site. With Google's much larger market share, the recent launch of its own travel site poses potentially far greater risks to competitors, to airlines, to competition in air travel bookings, and ultimately to consumers. How will Google ensure fair competition in travel bookings?
Demonstrating harm to consumers is essential to demonstrating the need for government action. The mere assertion of the possibility of harm is not sufficient to justify legal or regulatory intervention. However, the potential for harm cannot simply be dismissed without more information.
For example, Copiepresse, which represents Belgian newspapers, complained about Google's use of their content and found that they were no longer listed in Google search at all. The resulting loss of traffic eventually led them to cooperate.