Let's Play

The fact that publishers tailor their web pages to search engines and their algorithms can be a huge headache for the search engine providers. A search algorithm that worked well in the laboratory might deteriorate once it is let out in the wild. As a search engine becomes popular and publishers experiment with ways to skew search results in their favor, the general publishing behavior changes, conditions and assumptions which held in the laboratory no longer apply, and the search algorithm becomes a victim of its own success. I liken the situation to the problem an anthropologist studying a hitherto secluded, aboriginal tribe might have to contend with: the mere presence of the anthropologist in their midst, induces behavioral changes in the tribe that will taint the study.

Thus a successful search algorithm creates a feedback loop that modifies the behavior of web publishers. From the search engine provider's perspective, there are 2 broad approaches for dealing with this challenge:

  1. Ensure that the algorithm is impervious to modifications in publishing behavior.
  2. Modify or tweak the search algorithm as behavioral changes in the publishing activity emerge that threaten to break it.

Google, for example, takes the first approach: it ranks search results according to a page-weight scheme they call Page Rank. This weight algorithm is designed to be impervious to tinkering by web publishers. Of course, in practice, few things are impervious to the human will, and publishers are likely to find ways to "fake" a high Page Rank in order to skew Google's search algorithm in their favor. Once the "exploit" is widely used by publishers, you can expect Google to apply the second approach, above, namely to adjust its algorithm in an attempt to negate the exploit.

With the Ila project, however, no attempt is made to mitigate publishers' behavioral changes. Rather, the interface delegates this responsibility to the user. Ila provides a simple yet, I believe, sufficiently general way for users to filter and retrieve a subset of links. This capability allows users (or a link-search algorithm) to decide which types of links are relevant to a search. Quite the contrary, Ila is designed to encourage publishers to exploit it. That is, it begs authors to tailor their linking strategy for use by link aware users. Let's see how this might work.