Time to Eat My Words: The Search Box Grows

Half a year ago I wrote a blogpost about an easy change that Google could make to its interface, one that would both sacrifice only the least bit of simplicity as well as entice and encourage the user to enter longer queries, thus improving retrieval effectiveness.  In particular, I wrote:

So even though research has found that longer queries lead to more satisfied users, and that larger query input boxes lead to longer queries, Google is unable to take an evolutionary step in that direction.  That step violates their current locally-maximum hill principle of simplicity.  They seem fundamentally incapable of passing through the valley of complexity to reach an even higher effectiveness peak because evolutionary thinking does not allow them to take that large leap necessary.  They can only follow their current gradient.  In ten years of using Google, I don’t think that I have ever seen, even for brief experimental time periods, a query input box that was taller than one line.  Thus, evolutionary thinking conflicts with long-term goals.

Well, it’s time for me to eat those words.  For today, the search box grew in size.  From the official Google blog:

[S]tarting today, you’ll notice on our homepage and on our search results pages, our search box is growing in size. Although this is a very simple idea and an even simpler change, we’re excited about it — because it symbolizes our focus on search.

It is indeed a very simple idea, which is why I’ve been so stymied as to why the step wasn’t taken years ago.  The Belkin et al paper that I cite in my earlier post, the one that does the study on larger query input boxes leading to longer queries and more effective retrieval, was published at SIGIR 2003 — six years ago.  With 70% of Google’s workforce focused on search, you’d think that they’d have gotten to this one years ago.  Well, glad to see it has finally arrived.

I’m also curious about why, when enlarging the input box, they decided to make it wider rather than deeper.  Was this a result of an A/B test?  Did they try it both ways?  What did they look for in the logs, to tell them whether horizontal or vertical enlargement was better?

And I’m also curious: When they finally made the decision to even consider enlarging the input box in the first place (before they even starting figuring out the best way to do it), was it as a result of reading studies such as the Belkin one?  Or was there something that they found in the log analysis that told them that a larger input box would be better?

Having never directly worked in this area, this is an ongoing question that I have about web-scale information retrieval: I’ve heard that if a researcher is thinking about adding new functionality to web search engine, that researcher has to be able to argue from a log analysis that a particular behavior already exists, i.e. that a fundamental need is not being met.  Yet how can a particular behavior exist if the mechanism for executing (acting out) that behavior doesn’t exist?  It’s a chicken-egg thing.  It’s fairly obvious that people will enter more query terms when you give them more room to do so. But intuitive obviousness is not the same as log analysis evidence.  If you’ve never given people more room to enter more query terms, how do you know whether people will enter more query terms when given more room?  How does log analysis overcome this logical circularity?

Here’s another example, from my Music Information Retrieval past: Query-by-humming.  How can you tell whether there are a large number of users that want to search for music on the web by humming it into the search box?  Your search box isn’t actually capable of taking live-streaming audio as an input, so you actually have no instances in the logs of people actually taking the humming action.  Sure, I suppose you could approximate it by looking for text queries such as [song melody humming] and [find song hum].  But the scale (pun not intended) will still be off by unknown order(s) of magnitude: The number of people looking for sites at which you can hum your query is not of the same magnitude as the number of people who would hum that query directly into your search engine, were it immediately and explicitly available.  The mere existence of a new form of query input alters the scale of people’s behaviors.

I’m starting to drift off topic, so I’ll wrap up. I am still curious, though, about how this decision to finally enlarge the search box was arrived at, as well as why the horizontal was expanded rather than the vertical.  Again, however, I am glad to see that the step was finally taken.