Keep your laws off my Algorithm: Reply to the NYT Editorial on Google

The liberals at the New York Times are again calling for the regulation of something they don't understand: search algorithms.  Google, on the defensive, offered a hackneyed response, allegedly attributed to Marisa Mayer (analysis here from Tom Krazit at C|Net, since I don't have an FT subscription).

Concern about spammers gaming search results is valid, but, as a search insider (I work at Bing), I want to clear up some common misconceptions about the fabled search "algorithm" and offer some other salient reasons why regulators should stay away from search engines. If it feels like I'm repeating material from Death of the Search Startup, well... I am =)

A clip from the NYT editorial:

Still, the potential impact of Google’s algorithm on the Internet economy is such that it is worth exploring ways to ensure that the editorial policy guiding Google’s tweaks is solely intended to improve the quality of the results and not to help Google’s other businesses.

The NYT seems to be under the horrible misconception that there is an "editorial policy" guiding algorithm tweaks.  Misleading articles about Google's algorithm have propagated a viewpoint that there's a Google Editorial Committee lead by Udi Manber that looks at side-by-side search results and decides which looks better.  Nothing could be further from the truth.  To change the ranking of results, search engines use a potpourri of techniques from machine learning, to graph theory, to hand-tuning of features.  Remember that any change to the algorithm affects somewhere between thousands and millions of queries.  That means that if you measured improvement by looking at how a few results changed for the better, you might be ignoring potentially millions of results that got worse.  To measure the algorithm's progress, search engines use a complicated matrix of human judgments and click data, all analyzed by a bunch of smart PhDs and statisticians. 

Does anyone seriously think a government regulator is going to be able to add a lot of value to that process?

Now, a valid question might be: what if search engines are editorially changing their search results?  Neither Google nor Bing has a lot of content even to put in search results.  Beyond MSN and Knol, search engines just don't produce original content.

It is certainly true that SERPs (search engine result pages) are getting richer with many "answers."  I just did a a search for Britney Spears on Google and got the following 3 answers, which support three different Google properties.

As someone who works on the UX of the SERP, I can tell you two few things: 1) The coolest, most interesting, and most useful innovations coming up in search are going to be with rich content on the SERP and I'd hate for the government to stifle us 2) Search engines are incredibley worried about rich elements on the SERP that take users away from a satisfied click.

Imagine this: Google puts huge, garish maps/images/news/video answers on every SERP.  Either users will hate the new interface and abandon Google for a competitor OR the new elements will add value to the users and they'll enjoy the new experience.  Google would be shooting itself in the foot if it sent users to pages they didn't like, even it it provided a short-term bump in Google's overall page views.

My comments about the "algorithm" (I loathe to take it out of quotes) and answers suggest something inherent to search engines: we are naturally self-regulating because any changes made to the SERP or results will change user behavior.  Because search engines want to optimize for user satisfaction, there's no choice but to be agnostic to content that's shown on the page.  Therefore, I admonish regulators to keep out of search until there seems to be a demonstrable problem.

Of course, there's a great libertarian argument to keep the government's hands off search engines, but I won't go there and will simply let Reason speak for me.

Oh, and if you're worried about a Google monopoly, then try an alternative.  Simple as that.

Follow me on Twitter.

[Obligatory note: the opinions expressed above are mine alone and do not necessarily reflect the opinions of my employer, Microsoft.]

Comments

[Mark@Life Insurance]

I have read a blog a while ago and a guy was saying that he attended one of the meetings of Google and they were talking about search tactic and how different the results will be with just a single letter even the capitalization has effects.

[JB]

"The NYT seems to be under the horrible misconception that there is an "editorial policy" guiding algorithm tweaks."

Pure projection. They are keenly aware of their "editorial policy" guiding "news" tweaks.

[Regan Marye]

For a reputable institution, I can hardly believe that they've made that statement. I do however find it funny that people would want to regulate something that is so definite as an algorithm, and turn it into something biased. That sort of defeats its whole purpose.

[status sms]

I do however find it funny that people would want to regulate something that is so definite as an algorithm, and turn it into something biased. That sort of defeats its whole purpose.

[e urdu point]

That sort of defeats its whole purpose.