Pietro (fool_in_spirit) wrote in suggestions,
Pietro
fool_in_spirit
suggestions

LJ as a small world

Title
LJ as a small world

Short, concise description of the idea
Implement a search machine that given to users find the sortest path between them. Many things are then possible...

Full description of the idea
Read the first comment to read this paragraph correctly spaced. In practical details I suggest to implement a search machine that given to users x, y tells you the shortest path to go from x to y _and_ from y to x. I also suggest to use this search machine to test the average distance between users, the existance of a core of elements all interconnected (as many theories now suggest... no ref available at this time). The structure of this core: average distance, maximum distance, etc. Those info could be added to the stats page. The algorithm should be able to work between general users, but then should be possible for a logged in person, who reads another user to just see above, the informations on the distance between the person he is viewing and himself. Once this is done it will be possible to gather information on lj as a small world. Lj is giving a great possibility to test one of the current ideas of sociology. That is the idea of small world network. In brief a sw is a graph locally mainly similar to a lattice or ordered graph, but with some long distance wiring. This gives the possibility to elements that would be far apart to be actually few nodes away. Sw lie somwhere between random net and ordered ones, yet share some properties of both. The net of artists (who worked with whom) had been shown to be a small world. ecc. Now LJ is surely a small world. What we just need to do is to prove it. Not only, being quite a vast network, closely monitored we have the possibility to study some extra ideas. Example: We can surely divide lj users in 3 categories: a) a core of element interconnected, b) elements that can be reached from this core, but don't point back, c) element that point to this core (directly or indirectly) but can't be reached back. This has been shown to be the structure of the www in a recent paper. Now how is this core structured. I suspect that a similar pattern exists at this level, with a small set of users closely connected between each other (let's call this CENTER), and the other spreading out. To think about the CENTER think of users like brad, shannonkringen, gardengnome, people who have many friends and are considered friends by often even more people. Yet they don't all list each other. Not only, some people could have only few friends, but key ones. Listing those people is harder but I suspect with such search machine would be possible I suggest to implement the search machine, gather some data, look at the data (and publishy it in a web page). From then should be possible to write a paper on the subject.

An ordered list of benefits

  • Possibility for the average user to discover how far apart a person is. And if they are interested in this person who might be potential new friends.
  • Possibility to write an academic paper on the subject. Many people here are still getting a degree, having already a paper published can mean quite alot in terms of curriculum vitae.
  • Possibility to ask for MUCH MONEY from the USA government to study through LJ the structure of nets like internet. Around a million dollar. Before 11/9 the study of internet was one of the main interests for the US government, and I suspect it hasn't dropped that much. If, after having written a paper we want to ask for some money from the government this might be possible.
  • Possibility to use informations gathered to turn down abuses. (Sort of: I only accept comments from people who either are in the central core of lj or are signed as friends by one of those persons, similar to the trust metric, but way more simiple to implement).
  • Well, is actually quite interesting to use something like this to find
    a: who are thos core of lj.
    b:How are they connected.
    c:How does lj expands from them? Uniformly or with branches like a galaxy. (Personally I find this last point Very interesting).
    Is a bit like opening the eyes from above.
  • Possibility to know immediatly the random user wuo is he/she. This person have added me, how did this person probably found my lj? (of course to use this feature we need a button that asks the shortest pathe between user x and user y, except the direct path xy or yx )
  • Possibility to see a random user we found through http://www.livejournal.com/random.bml
    how would naturally relate with us.
  • An ordered list of problems/issues involved

  • Well, you know the mentality of villages: who do you know?
    Hmm you are a stranger. Well, here could make lj even more have this mentality: hmm you are not in the CENTER, I don't speak with you.
  • Work in implementing it.
  • Gosh, another bloody paper.
  • Server load? Especially if this feature would be used often (like in automatic). Yet, if not I don't think it would charge so much.
  • An organized list, or a few short paragraphs detailing suggestions for implementation

  • First should be implemented the general search machine. This should not be a big deal.
    something like:
    List Search (user x, user y);
    Well, I don't know pearl, sorry.
    Then should be made available to users maybe in S2.
  • I am availale to discuss any extra clarification needed.
  • Tags: searches, § implemented
    Subscribe
    • Post a new comment

      Error

      Anonymous comments are disabled in this journal

      default userpic

      Your reply will be screened

      Your IP address will be recorded 

    • 12 comments