Sunday, December 05, 2010

The Case for Social Search

An interesting and well written article in the NY Times about the owner of the online shop decormyeyes caught everyone's attention: the owner actually admits to willfully create negative ratings for his shop by actually giving crappy customer service... because this actually pushes him up in Google search results, due to the nature of the page rank algorithm. And thus the bad reputation actually brings him additional business.

Because everytime someone has a problem with him and posts this to a blog or a consumer feedback site like there is a link to his site (with a mention of the product they purchased), and each of those links counts in the PageRank alogrithm, which is primarily based on the number of incoming links to a site. The more links point to your page, the more relevant this page is considered.
Due to the large number of complaints, he actually made #1 for many of the goods he sold on his site.
And shoppers searching for the item were likely to shop there, because it was #1 search result. Organic search result - that is - not a paid ad.

No-one saw all the negative feedback, because Google does not show it (or actually know about it).
However, if you do a search on the site's name you will also see a lot of complaints about decormyeyes on the first results page. So searching for the site/shop does work as expected.

Odd that it took about 10 years for anyone to notice this flaw...

The problem there is, that Google search results are not recursive... i.e. you do not see the results of a result.
(Maybe this would actually fix the problem).

Google is not to blame for this, and the PageRank algorithm is still fine. If Google were to implement some semantic magic in their algorithm that would treat a positive reference different from a negative reference, it would create an altogether different problem: for all the controversial topics (or persons) out in the world, you simply will always find negative and positive references, and in many cases the negative will outnumber the positive. The system would be gamed with as easily as the current.

That's what Google said in reaction to this story:
As it turns out, Google has a world-class sentiment analysis system (Large-Scale Sentiment Analysis for News and Blogs). But if we demoted web pages that have negative comments against them, you might not be able to find information about many elected officials, not to mention a lot of important but controversial concepts. So far we have not found an effective way to significantly improve search using sentiment analysis. Of course, we will continue trying.
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.
In other words, for the time being, Google will - on a very high level - curate the search results.

What worries me (if only slightly) here is that Google will use "their opinion" to determine which merchants (in this case) to filter.

Although I consider this is the right solution for now, it has two flaws:
  1. It can only work a posteriori,
    i.e. only after you find the next merchant (or content provider, or ...) who games the system like this, you can remove him from search results, but this means that they already did damage.
  2. This is the by-far #1 search engine - let's call them a monopoly for the sake of this argument - applying their opinion/judgement....worrisome, isn't it ?

The solution:

Take the opinion out of Google's hands. Crowd-source the opinion.

Let the wisdom of crowds or rather the opinion of crowds do the work.

Which crowd? My crowd of course. For me.
Your crowd for you.

In other words: Social Search

So the decormyeyes incident - apart from being disgusting from a customer care perspective - can finally help make the case for social search.

Did you notice that in this entire post I never linked to the actual site, just to avoid contributing to the problem ;-)

No comments: