Otis, you might be right that their previous related searches were strictly based on text-wise query similarity. Too late to check now! But I think that’s still largely true of the current ones.

For example, compare:

http://www.google.com/search?q=sigir

http://www.kosmix.com/topic/sigir

And, in case anyone is wondering, I have no vested interest in Kosmix, and more than in Duck Duck Go or any other web search engine. But I do make a point of plugging the folks who are pushing the envelope of exploratory search on the web.

Oh my goodness — don’t even get me started. I agree, that is a strange, strange assumption. It’s almost completely backwards. The statistical distribution of eye movement across the page is not independent of the design itself. Change the design, and you’ll change the golden triangle into an oval, square, vertical striped lines, whatever. It’s almost as if they assume that the golden triangle is an objective, decontextualized fundamental law of the cosmos, and design everything to fit into that. That doesn’t make any sense to me.

(FWIW, it might have been Yahoo, but Google also does this golden triangle stuff:

http://googleblog.blogspot.com/2009/02/eye-tracking-studies-more-than-meets.html)

About your CHI workshop paper: That sounds quite interesting! Any way you could send me a copy? Or do you have to wait until it’s officially published?

Jeremy – we are about to promote this same position about getting context from facets without specifically interacting with them, at the sensemaking workshop at CHI. It follows on from our Backward Highlighting paper at UIST, which did try to measure it to some extent.

One could read a whole lot into that.

I’ve heard them comment about this. They say that the reason why they put things like this at the bottom, only, is that they don’t want it to interfere with your normal successful searching.

Their opinion, if I understand it correctly, is that if a query is a successful query, and you find exactly what you want in the 1st or 2nd rank, then Google’s opinion is that having the facets at the top or at the side only gets in the way of you getting to that top result as quickly as possible.

It is only when the query is an unsuccessful query that you start scrolling down further in the ranked list, so that you eventually wind up at the bottom of the list. At that point, you see the query suggestions, and only at that point do they become relevant to you, i.e. when your query has failed.

This is Google’s story, and they appear to be sticking to it.

Now, on the surface it sounds reasonable. But I don’t believe it, for two reasons:

(1) Long ago they started putting ads above the top-ranked result, instead of just over to the side at the right. So they already throw up junk in front of the top ranked search result, and make it harder for the user to find/click that top result.

(2) As a searcher, I don’t always use exploratory facets or query suggestions in an explicit way. Often I use them to conceptually orient myself in the searches that I do. I read over the facets to get an understanding of what the space is about, without actually clicking on them. They aren’t always directly useful, but they are extremely valuable in helping me establish a context around the rest of the information that I am seeing.

Of course, Google doesn’t realize this, because they can’t measure it. If nobody clicks one of the query suggestions or facets, they have no way of knowing whether those facets were useful.. even if they were extremely useful to me.

And so having facets at the top, even if I never click them, doesn’t interfere with my normal, successful searching, anyway, as Google fears. Even if I never click anything, seeing that information is valuable to me.