Monday, April 21, 2008

The Rift between Industry and Academia?

The gulf of difference in the research approach in the industry and the academia is markedly astounding. Recently there was this article on the Wired blog about how Google is losing grounds in advertising with the new Web 2.0 tools coming up. The article is here. It describes how, even with the rich and extensive user data as well as data about a person's online social activities (e.g. Orkut) Google has been sticking to its conventional "search" ideas to targeted advertising and therefore losing grounds monetarily.

I work on social networks broadly and my research deals tremendous impact on tomorrow's new definition of targeted advertising. So I am rather dumb-founded finding out how Google is targeting advertising in a social media simply based on analysis of the textual content. This might have several limitations:
  1. On a social media, the content on a person's page could be generated by the person or by his contacts or even by people who aren't related to the person in any way. Targeted advertising focused on the person and based on the textual content would therefore fail.
  2. The person might receive a lot of spam messages / scraps (note, certain people on a social network could be spam, which a conventional spam detection algorithm can't possibly find out; "splog detection" is a separate research area therefore). However the context of these spam is no way related to the user.
  3. In a very natural setting of the society, people acquire roles in communication. When I go a party, I know that Jane is the girl most outgoing and likely to know loads of people - a role culling out of Jane's interactions with people. Why not target such people for targeted advertising?
  4. The article apparently talks of a small company like a start-up which is exploiting this idea of Influencers in the social networking advertising domain. However, they are missing out on another interesting attribute of people in a social network - that some of us are more susceptible to certain kinds of information than others. Some of us are inherently vulnerable to the influencers.
This is the strong point of surprise and rift with the academia. We have known from simulations (ref. Duncan Watts) that flow of information in a social network depends on mining the vulnerable people as well. And through our work on MySpace (still ongoing work for a journal), we want to establish the empirical validity as well. But the interesting issue is, is the industry not aware of this powerful idea? Or they don't really care about it?

Google with its million Terrabytes of data, could actually work wonders in realizing a dream of putting social networks to very strong use beyond the commonplace "friend-of-friend" notion!

No comments: