Much is often made of how Google could improve search results (or do a lot of evil) by learning from our search histories. Since there are a lot of valid privacy concerns with that approach, I think Google could still improve results considerably by learning from a user’s result-clicks within a search session.

As an example, I Googled “Roll your own” earlier today, looking for white label web solutions. The results included a lot of “roll your own cigarette”  items. Given that I ignored all of those and instead clicked on the tech-related results, each time I returned to Google using the back button and then moved to the next results page, Google had an opportunity to filter the results on that next page to remove items that I didn’t seem interested in.

I guess the Similar Items feature is meant to do this sort of filtering but for some reason I never remember to use it.

So what do you think - would you find this kind of feature useful?