Thursday, December 02, 2010

Google admits its algorithm is opinion, but its decision process is dangerous

In an online discussion just after Web 2.0 Summit Executive Producer John Battelle's blog post on how Google staff responded to the problem presented by the bad business man who took negative comments about his business and turned them into fuel for positive search rankings, Google Sofware Engineer Matt Cutts wrote a comment that scarily reveals how the search firm goes about constructing its search algorithms, and reveals what this blogger has asserted for years: that they are just the views of a group of people, not some kind of fact.

Or, to think of it another way, before we continue, Google's "opinion" as expressed by its algorithms, is that blogs should be pushed to second-tier search status (apparently regardless of the income impact on webmasters) that The Associated Press' view matter more than that of TMZ.com, and online speech expressed in a certain way and by a handful of media sites, is what you should see first on Google News. In short, Google's algorithms, both in design and development process, take the look of an instrument of anti-democracy in the wake of Matt Cutts' comments.

That's certainly the case with how the Google News Meta Tags Program has been implemented, and its result: sacking hundreds, if not thousands of blogs and website from Google News under the cover reason of "spammy sites," while really protecting Big Media in the form of the AP.

John Battelle's Blog; Matt Cutts Take

In his blog, Battelle rightly points to a rather disturbing Google statement in its blog regarding how the firm's staff collectively responded to the problem presented by the New York-based bad business guy. This one:

Even though our initial analysis pointed to this being an edge case and not a widespread problem in our search results, we immediately convened a team that looked carefully at the issue. That team developed an initial algorithmic solution, implemented it, and the solution is already live...Instead, in the last few days we developed an algorithmic solution which detects the merchant from the Times article along with hundreds of other merchants that, in our opinion, provide an extremely poor user experience. The algorithm we incorporated into our search rankings represents an initial solution to this issue, and Google users are now getting a better experience as a result.

John locked on the use of one word: "opinion." The use of that word surprised him, as John wrote "If ever there was an argument that algorithms are subjective, there it is."

This blogger find not just the use of the word "opinion" bothersome but Google's admission that it focused on the resolution of a problem caused by one person, and based on a New York Times article to be extremely disturbing. Given the size of the impact of changes in Google's search algorithms on the lives and incomes of millions of people, Google should have employed the same system dynamics and scenario planning approaches that the World' largest oil companies use in their decision making.

While Google looked at how others were behaving in search, and just to make sure it was not a "wide spread problem" in the Google-World, Google did not ask how its changed focused on one person would possibly negatively impact others who use their search engine - and that's most of the industrialized World.

Oil companies are painfully aware of the World-altering impact of their decisions, so much so that modeling paradigms which cause them to take "a big World view" are used. That's obviously not true for Google, yet the search company arguably has the same level of potential impact on the World. Want more proof of Google's apparent failure to see the large-scale impact of its actions? Take a look at what Matt Cutts wrote on John's blog:


The latter commenters are getting it right, but I believe the "opinion" in that sentence refers to the fact our web search results are protected speech in the First Amendment sense. Court cases in the U.S. (search for SearchKing or Kinderstart) have ruled that Google's search results are opinion.
This particular situation serves to demonstrate that fact: Google decided to write an algorithm to tackle the issue reported in the New York Times. We chose which signals to incorporate and how to blend them. Ultimately, although the results that emerge from that process are algorithmic, I would absolutely defend that they're also our opinion as well, not some mathematically objective truth.
Here's how I phrased it in 2006 in an interview with John that appeared on this site: "When savvy people think about Google, they think about algorithms, and algorithms are an important part of Google. But algorithms aren't magic; they don't leap fully-formed from computers like Athena bursting from the head of Zeus. Algorithms are written by people. People have to decide the starting points and inputs to algorithms. And quite often, those inputs are based on human contributions in some way."


And it's the way Google frames those "human contributions" that's to be questioned. Again, the way Google comes to these decisions is totally out of scale to the enormous potential impact they have.

Want proof? Look at the bloggers and webmasters kicked off Google News, and insultingly called spammy sites, when many are not, while giants content farms like Examiner.com and Associated Content that do spam Google News are allowed to remain, and Google makes deals to help The Associated Press , and not blogs, seek new revenue areas (true and stated by Google's Senior Business Product Manager, Josh Cohen).

Matt Cutts Elaborates On Google's Opinion

Cutts then writes yet another comment on the same blog posts, that's even more revealing and backs my assertion that Google does not conduct the proper risk analysis in considering the overall impact of the changes it makes to its search algorithm. Moreover, Cutts final statement is even more bothersome:


"I doubt very much that this algorithm was developed as a response to NY Times."
Anthony, I had a ringside seat, so I know it was. We still did tons of testing (e.g. running hundreds of thousands of queries to see changes), but the new algorithm happened in response to that article.
Mor, you raise an important point too: "Interestingly, some time between 2002 and 2009 Google dropped that reference from their corporate "tech page". "PageRank performs an objective measurement..." became "PageRank reflects our view of the importance...". "
I had a ringside seat for that one too; here's my informal/personal summary. SearchKing sued Google and the resulting court case ruled that Google's actions were protected under the first amendment.
Later, KinderStart sued Google. You would think that the SearchKing case would cover the issue, but part of KinderStart's argument was that Google talked about the mathematical aspects of PageRank in our website documentation. KinderStart not only lost that lawsuit, but KinderStart's lawyer was sanctioned for making claims he couldn't back up. See http://blog.ericgoldman.org/archives/2007/03/kinderstart_v_g_2.htm for more info.
After the KinderStart lawsuit, we went through our website documentation. Even though Google won the case, we tried to clarify where possible that although we employ algorithms in our rankings, ultimately we consider our search results to be our opinion.
For example, what should rank number one for the query [barack obama]? Obama's personal website, or the White House? What should rank #2, or #10? The fact is, there's no objectively "correct" way to rank those results--reasonable people can disagree whether Obama's Twitter or Facebook page should rank higher than (say) a Chicago Tribune article about Obama.
That single point, which courts have agreed with, proves that there's no universally agreed-upon way to rank search results in response to a query. Therefore, web rankings (even if generated by an algorithm) are are an expression of that search engine's particular philosophy.
In this case, Google chose to dedicate engineers to write a new algorithm that we believe improves our web rankings. If you consider fledgling search engines such as Blekko and DuckDuckGo, they chose to remove the Decor My Eyes site completely--you can't find the website even if you search for the exact url by name on their sites. That's their philosophical choice--their opinion of how best to handle such searches, and I support their decision. I think each search engine needs to be free to rank results in the way that they think is best, and if people believe one search engine returns better results, it's easy to switch. That's how Google went from 0% market share to its current position.

Here's more proof that Matt Cutts fails to understand Google's current World impact, and that Goggle staff are perhaps too much insular than is healthy for a firm that has its overall market share and influence.   Cutts basically says if you don't like Google, go somewhere else.   That's the problem: because most of the industrialized World not only goes to Google, but is directed to use Google by web browser partnerships, going elsewhere is improbable.  Google's brand is such a part of our culture that its company name has become a verb.  Matt's statements, and Google's and Google News' actions, do not reflect a company that understands that.  

Google must install the proper risk assessment decision systems.  Perhaps doing so would keep them out of court as much as they have been in it, and help Google understand that it's role is to help the World, not the World's elite media.

No comments:

Post a Comment