Monthly Archives: September 2011
Over the last couple of weeks, things have heated up considerably for Google – on the mobile side with the start of a patent war, but also in the search area, the core of the company’s business. Led by Senator Mike Lee (a Utah Republican), the US Senate’s Antitrust Subcommittee has started to probe into certain aspects of Google’s ranking mechanisms and potential cases of abuse and manipulation.
In a hearing on Wednesday, Lee confronted Eric Schmidt with accusations of tampering with results and the evidence the Senator presented was in fact very interesting because it raises the question of how to show or even prove that a highly complex algorithmic procedure “has been tampered with”. As you can see in this video, a scatter-plot from an “independent study” that compares the search ranking for three price comparison sites (Nextag, Pricegrabber, and Shopper) with Google Price Search using 650 shopping related queries. What we can see on the graph is that while there is considerable variation in ranking for the competitors (a site shows up first for one query and way down for another), Google’s site seems to consistently stick to place three. Lee makes this astounding difference the core of his argument and directly asks Schmidt: “These results are in fact the result of the same algorithm as the rankings for the other comparison sites?” The answer is interesting in itself as Schmidt argues that Google’s service is not a product comparison site but a “product site” and that the study basically compares apples to oranges (“they are different animals”). Lee then homes in on the “uncanny” statistical regularity and says “I don’t know whether you call this a separate algorithm or whether you’re reverse engineered a single algorithm, but either way, you’ve cooked it!” to which Schmidt replies “I can assure you that we haven’t cooked anything.”
According to this LA Times article, Schmidt’s testimony did not satisfy the senators and there’s open talk about bias and conflict of interest. I would like to add to add three things here:
1) The debate shows a real mismatch between 20th century concepts of both bias and technology and the 21st century challenge to both of these question that comes in the form of Google. For the senator, bias is something very blatant and obvious, a malicious individual going to the server room at night, tempering with the machinery, transforming the pure technological objectivity into travesty by inserting a line of code that puts Google to third place most of the time. The problem with this view is of course that it makes a clear and strong distinction between a “biased” and an “unbiased” algorithm and clearly misses the point that every ranking procedure implies a bias. If Schmidt says “We haven’t cooked anything!”, who has written the algorithm? If it comes to an audit of Google’s code, I am certain that no “smoking gun” in the form of a primitive and obvious “manipulation” will be found. If Google wants to favor its own services, there are much more subtle and efficient ways to do so – the company does have the best SEO team one could possibly imagine after all. There is simply no need to “cook” anything if you are the one who specifies the features of the algorithm.
2) The research method applied in the mentioned study however is really quite interesting and I am curious to see how far the Senate committee will be able to take the argument. The statistical regularity shown is certainly astounding and if the hearings attain a deeper level of technological expertise, Google may be forced to detail a significant portion of its ranking procedures to show how something like this can happen. It would, of course, be extremely simple to break the pattern by introducing some random element that does not affect the average rank but adds variation. That’s also the reason why I think that Lee’s argument will ultimately fizzle.
3) The core of the problem, I would argue, is not so much the question of manipulation but the fact that by branching into more and more commercial areas, Google finds itself in a market configuration where conflicts of interest are popping up everywhere they turn. As both a search business and an actor on many of the markets that are, at least in part, ordered by the visibility layering in search results, there is a fundamental and structural problem that cannot be solved by any kind of imagined technical neutrality. Even if there is no “in house SEO” going on, the mere fact that Google search prominently links to other company services could already be seen as problematic. In a sense, Senator Lee’s argument actually creates a potentially useful “way out”: if there is no evil line of code written in the dark of night, no “smoking gun”, then everything is fine. The systematic conflict of interest persists however, and I do not believe that more subtle forms of bias towards Google services could be proven or even be seriously debated in a court of law. This level of technicality, I would argue, is no longer (fully) in reach for this kind of causal demonstration. Not so much because of the complexity of the algorithms, but rather because the “state” of the machine includes the full structure of the dataset it is working on, which means the full index in this case. To understand what Google’s algorithms actually do, looking at these algorithms without the data is no longer enough. And the data is big. Very big.
As you can see, I am quite pessimistic about the possibility to bring the kind of argumentation presented by Senator Lee to a real conclusion. If the case against Microsoft is an indicator, I would argue that this pessimism is warranted.
I do believe that we need to concentrate much more on the principal conflicts of interest rather than actual cases of abuse that may be simply too difficult to prove. The fundamental question is really how far a search company that controls such a large portion of the global market should be allowed to be active in other markets. And, really, should a single company control the search market in the first place? Limiting the very potential for abuse is, in my view, the road that legislators and regulators should take, rather than picking a fight over technological issues that they simply cannot win in the long run.
EDIT: Google has compiled its own Guide to the Hearing. Interesting.
German publisher Heise Verlag is an international curiosity. It publishes a small number of highly influential computer-related magazines that give a voice to a tech ethos that is at the same time extremely competent in the subject matter (I’ve been a steady subscriber to c’t magazin for over 15 years now, and I am still baffled sometimes just how good it is) and very much aware of the social and political implications of computing (their online magazine Telepolis testifies to that).
Data protection and privacy are long-standing concerns of the heise editors and true to a spirit of society-oriented design, they have introduced a concept as well as a technical implementation of a two-step “like” button. Such buttons, by Facebook or other companies, have of course become a major vector of user-tracking on the Web. By using an iframe, every button loads some code from Facebook’s server and sends the referring url (e.g. http://nytimes.com/articlename/blabla) as an information. The iframe being hosted on the facebook.com domain, cross-site privacy protections can be circumvented, the url information connected to an identifier cookie and, consequently, to a user account. Plugins like the Priv3 project block these mechanisms but a) users have to have a heightened level of awareness to even consider installing something like this and b) the plugin interferes with convenient functions like Google search preferences.
Heise’s suggestion, which they already implemented on their own sites, is simple: websites can download a small bit of code that implements a two-step procedure: the “like” button is greyed out after the page first loads and there is no tracking happening. A first click on the button loads the “real” Facebook code, and the second click provides the usual functionality. The solution is very simple to implement and really a very minor inconvenience. Independently from the debate whether “like” buttons and such add any real value to the Web, this example shows that “social” features like these can be designed in a way that does not necessarily lead to pervasive user tracking.
The echo to this initiative has been very strong (check the Slashdot discussion here), especially in Germany, where privacy (or rather Datenschutz, a concept less centered on the individual but rather on the role of data in society) is an intensely debated issue, due to obvious historical reasons. Facebook apparently threatened to blacklist heise.de at a point, but has since then backpedaled. After all, c’t magazin prints around 600.000 issues of every number and is extremely influential in the German (and Dutch!) computer landscape. I am very curious to see how this story unfolds, because let’s be clear: Facebook’s earning potential is closely tied to its capacity to capture, enrich, and analyze user data.
This initiative – and the Heise ethos in general – underscores that a “respectable” and sober engineering culture does not exclude an explicit normative stance on social and political issues. And is shows that this stance can be translated into technical models, implemented, and shared, both as an idea and as code.