Imagine a search engine that employs experiments to learn from you? The computer scientists at Cornell University have and they are using stimulus money (a four-year, $1 million grant) to improve search based on such experiments. The scientists are looking to develop search engine software that can read your queries and clicks, as well as subsequent query reformulations, in order to understand search methodology and what does and doesn’t work. The software developed from these efforts will be best used in specialized collections – the examples from the press release include scientific and legal collections and corporate intranets.
The search engine software will learn what works best by analyzing user data, almost as if by osmosis. The researchers already have developed the aptly-named Osmot and are looking to improve the process by tightening the experimental controls. More on Osmot at this link.
What does it all mean for you? Smarter search engines might yield faster and better results, but I still hesitate slightly at the thought of a machine’s judgment regarding what is and is not relevant substituting for my own. In any event, it will be interesting to see where this inqury leads and the ramifications for search.