Decentralized search engines?

This post is more of a question than presenting a topic. In class this week we have been learning about decentralized searches and how people in social networks can quickly find the shortest path because of scales of resolution. I’m wondering whether the same idea can be applied to searching the web. Google does a great job of navigating the Internet, but it’s PageRank algorithm doesn’t exactly utilize the inherent structure of the Internet in a decentralized way. So my question is this: can a search algorithm be developed to find the shortest path between a user’s query and the target search results by using the decentralized search math we learned this week? Would this allow users to find more relevant results to their queries?

http://library.cern.ch/HEPLW/8/papers/1/

This is kind of confusing, but it's relevant. These researchers are attempting to do just that: build a decentralized search engine.

"The goal of the project is to build a tool capable of searching and automatically categorizing vast amounts of geographically distributed information, which in many cases could not be searched effectively, if at all, by a centralized search engine such as those available on the web."

The researchers are trying to use a grid system that resembles the social network grids that professor Immorlica explained in class.

Efficient Decentralised searching is the toughest task facing programmers today.

http://www.tech-faq.com/swarm-intelligence.shtml

Swarm intelligence seems to be the best theoretical way of searching - and google seems to use it tangentially in conducting its searching. I think that if a computer system could do decentralised searching without flooding a network with queries efficiently, it would have to resemble human intelligence. I can point to the homework question as an example where we had to find a path from one word to a target word. It was difficult! Can a computer find the most efficient route? I had to think ahead several steps and take into account several "random" links as explained in class.

I would think the best app for a thing like facebook would be one where you type in a start person and a target person and ask the app to find a most efficient route between the two.