A Speedy Rant on the Pitfalls of Modern Search
See newsletter title, oh my god
Search engines index pages by content-similarity rather than functionality. This is good for reference and focused retrieval queries, but very bad for exploratory queries, which make up the majority of many people’s search activity. Dumb! Oh my god!
When we use content-similarity to rank our results, we by definition keep relevant material locked behind a keyword that the user may or may not possess. When the user is performing the search as an exploratory activity, the likelihood that they already have access to one such keyword is not very high. Do you see where I’m going with this?
When we use content-similarity to rank our results, our results target content that is relevant to the phrasing of our query rather than to the intent of our query. Unless we happen to get lucky, we will not gain access to the relevant keywords that will allow us to perform the sort of focused retrieval query that defines the “success” of modern search engines.
In conversation on the birdsite recently, a few of us explored the role of context in the search process. I’ve been working with this particular problem for a while now, leading to the design of a function-first search experience that will (theoretically, *crosses fingers*) be better, but I’m shocked that I had to. This problem isn’t new, and while I do have more info rep background than the average person, anyone reasonably intelligent who’s ever had to play games with Google is capable of coming to the same conclusion.
Function over form, always and forever, in everything, but this especially.
Some fun questions for the future:
How might we infer relative domain ignorance from a query pattern?
Does the linguistic signature of domain ignorance look similar across agents?
What have we done to model semantic depth of a single keyword? Is there room for improvement?
What does that “better” search experience look like, anyway?
Coming soon :)