Tuesday, September 19, 2006

Trying to improve search user interfaces

Danny Sullivan has a good post, "Why Search Sucks & You Won't Fix It The Way You Think", that looks at the lack of innovation in search user interfaces and several failed attempts to improve them.

Danny does not get into details about why the attempts at improvements did not succeed, but I think it comes down to too much work.

Type stuff in a box, punch a button. That's easy and it works most of the time.

To be an improvement, any substantial change in that UI paradigm will have to be clearly easier to use or produce substantially better results almost all the time. Probably both. That's a high bar.

Type stuff, push a button. It's going to be hard to beat.

3 comments:

Anonymous said...

I generally like the things Danny has to say. But in this case I don't think he looked closely enough at the research on eliciting user feedback, e.g. query expansion suggestions provided by systems such as Ask, Altavista, and, from time to time, Yahoo.

I know I argue this point with you every single time you bring it up. So rather than throwing out hypotheticals, like I usually do, I will refer you to scientific studies in this area:

http://www.ischool.berkeley.edu/~hearst/irbook/10/node8.html

In particular, turn your attention to section 2: "The 64 subjects were much more effective...with relevance feedback than without it. The penetrable group [analogous to Ask.com] performed significantly better than the control [analogous to Google]...Search times did not differ significantly among the conditions, but there were significant differences in the number of feedback iterations. The subjects in the penetrable group required significantly fewer iterations to achieve better queries (an average of 5.8 cycles in the penetrable group, 8.2 cycles in the control group...)...All subjects preferred relevance feedback over the baseline system, and several remarked that they preferred the `lazy' approach of selecting suggested terms over having to think up their own."

When you say that users type something and push a button, you are oversimplifying the matter. When the query works on the first shot, that is great. But when the query does not work, and you can't find what you are looking for, users on Google have to manually go in and think up new terms to add, type those in, and push the button again. In order to meet their information need, users have to do this multiple times.

And, according to this summary by Marti Hearst, above, users took an average of 8.2 iterations when they had to think up their own query terms. However, they only took 5.8 iterations to find what they needed, when the system interface offered them Ask.com-like query expansion term suggestions.

Furthermore, users preferred the query suggestions (Ask's philosophy) over the manual approach (Google's philosophy).

The beauty in all this is that, when the user does not need to iterate, because the query worked the first time, he does not have to. Expansion suggestions are presented off to the side, and if the query worked, there is no reason to even look at those terms. If Google implemented this, it would be guaranteed to do no worse than it currently does. But for those queries where Google currently is not working very well (and we've all experienced that), there is the potential for huge, significant improvements in the number of iterations and amount of effort required from the user.

I think this study above really shows it.

So I completely agree with you that the user is lazy, and does not want to expend extra effort. But once again, I have to make the point that the Google approach is actually more effort-requiring than the Ask approach.

Greg Linden said...

That's a fair point, Jeremy.

I am guilty here of treating search as a one-shot deal. I agree it often is more like a dialogue where the searcher refines a query multiple times.

Moreover, I have other posts on this weblog talking about how search needs to move past the one-shot deal and start taking advantage of the fact that it is often a dialogue.

Thanks for keeping me honest on that. I think you are right that I was oversimplifying things on this one.

Anonymous said...

Goodness, you're too humble, Greg! :-) I wouldn't call you "guilty" of anything.. it was more the fact that I was trying to hammer out a number of disparate and often competing ideas. What does "simplicity" really mean? Is Google more simple, because of its clean interface? Or less simple, because of the lack of feedback mechanisms? Does user laziness mean they're too lazy to actually use a feature, or just too lazy to learn what that feature does? What if learning that feature saves them time/effort/etc in the long run? For example, as a child you could have been too lazy to learn how to walk, and spent the rest of your life crawling. But a little bit of effort, at the outset, has meant that you have saved lots of time and effort, in the long run. (Not to mention the fact that you'd look pretty silly :-) What does it mean, then, for a user to be lazy?

These are questions for which I think there is no easy answer. Google doesn't have it right. Ask doesn't have it right. They're tough questions. And I'm just trying to engage you on them.

Oh, by the way, I think one can actually characterize Findory, esp. the personalization, as a form of multiple-shot dialogue. It is more implicit, rather than explicit. But it is very much a step in that direction. So I realize you are thinking about these things.