And what about an algorithm that evolves by itself ?
Let me clarify about that:
- If we allow users to rate results (even passively), this enables a "collaborative filtering algorithm" that permit a dynamic reordering of results.
- This pseudo algorithm doesn't need to be updated, as it is based on users, and "you can fool a thousand time an algorithm, but you can't fool a thousand time a thousand of users": Humans won't be fooled each time, they are adaptive.
Actually, it would be great to allow users to submit by themselves new results, then let other users do the triaging job, to finally know if it was good stuff or not.
But the main problem, it has been said, is that:
- almost the whole webternet is indexed by few companies
- access to information for users to the whole webternet is controlled by the same few companies.
indeed, a part of answer can be found with the http://seeks-project.info/ that claims to do this kind stuff, over a decentralized open architecture, with a freesoftware p2p application. (beware, work in progress)
