Google: Algorithms Aren't the Only Answer

13 Comments

Google (s goog) has finally pulled back the curtain on a new feature that until now has been in restricted beta: the addition of wiki-style functions in standard search results. Once logged into a Google account, this allows you to click a small up or down arrow to move a specific result, click and delete it from your search entirely, or click on a small comment bubble and leave your comments on that result. Google will remember those settings the next time you search for the same keywords, and has said it may even work for similar or related searches. In many ways, Google is taking the same principles that power a site like Digg and applying them to search.

google_wiki480

Adding these kinds of features isn’t a universally popular move. When Wikia Search — Wikipedia founder Jimmy Wales’s attempt to do the same thing to search, with editing of results and comments (or “annotations”) encouraged — launched earlier this year, there was plenty of criticism aimed not just at the execution and the lack of usable results, but at the very concept of wiki-style search. Many said that opening search results up in such a way would leave the system vulnerable to the inevitable SEO gaming and trick-playing that hampers many other “crowd-sourced” services such as Digg.

This is a little like complaining that the furnace heating your house is too hot, and that you’re afraid it might burn someone. In many ways, wiki-style search is just an extension of the way that Google has always worked: that is, by aggregating the choices of millions of users and then using the PageRank algorithm to produce something approaching the best result. Voting and commenting features simply give Google more pieces of data they can use to arrive at the best result). They will also provide a fairly rich trove of activity-based information that the search engine could use to improve its regular results — that is, the ones that users who aren’t logged in will see — or to tweak its overall search algorithms based on the behaviour of wiki-search users. Why did so many people move that result up? Why did they move another down? Why did some delete that result and not others?

Will these new wiki-style functions be subject to rampant gaming and manipulation? Of course they will — just like everything else that the search giant touches. When you wield as much power online as Google does, gaming and manipulation follow in your wake like pilot fish following a shark. Presumably, the company has taken that into account, and will use their resources to reduce gaming as much as possible. And meanwhile, they will use the results of all that clicking to teach their engine a thing or two about human search behavior.

13 Comments

voip

This is more like a PeopleRank, and it will help Google to get insight on search results as long as they can hire the same amount of people reviewing the rankings by users. Nice try but I still believe in the good old PageRank, yes it does return poor results from time to time but it is still more trustworthy.

Engago Team

If ever Google starts using the results from Google Wiki for their normal searches, they will just kill themselves as Adsense has no value anymore.

Then a company just hires a company (India or China or Pakistan) having their employees promoting all the website links on searches.

Could become a nice business fro low cost labor countries and would put all SEO comapnies out of business too;

Markus Göbel's Tech News Comments

If the changes affect only my own searches, why would I want to make these wiki votes? Just in case for my next search on the same subject? That’s silly.

I am too lazy for this Web 2.0 stuff anyway. But I heard that Google uses its bookmark service to finetune its search results with human help. If that’s true, why wouldn’t they use these wiki-style functions too?

rohit

I think you are making an assumption that simple majority polling or ‘counting’ is used in this system. There are many different algorithms that can be utilized here and given the collection of data mining folks at that company, I would presume that simple ‘crowd sourcing’ is not what they are trying to fine tune…

Niraj

When you have the amount of data that they have coming in, it makes it a lot easier to prevent gaming the system. You can check for things like how many items a person has rated (those that game would be trying to influence specific results, not a wide range), how many people are providing ratings and if this same group of people is influencing results, etc. Much like the system Digg has in place to prevent gaming.

The real downside to all this is how ugly it makes search result pages look. They need to shrink the icons or make them less conspicuous. With three fairly large icons per result, it really looks cluttered right now.

Comments are closed.