Friday, May 27, 2005

Yahoo MindSet from Yahoo Research

Bernard Mangold has the announcement on the Yahoo Search blog about Yahoo MindSet, a prototype that reorders search results "according to whether they are more commercial or more informational" as you move a little slider around.

It's pretty similar to but more limited than Google Labs Personalized Search, which allows searchers to check boxes for their high level subject interests and then gives you a slider to reorder search results.

MSN Search also has some similar sliders (though they don't reorder dynamically) on their main search site inside their Search Builder advanced interface.

I'm surprised to see this focus on sliders. They aren't particularly useful. They fail the grandma test. Most novice users will not use or understand the slider; they just want the top result to be useful. It's not even that useful to power users since sliders fail to provide the level of granularity they need.

More knobs, more buttons, more complexity. With less than 1% of users even bothering with the existing advanced search options on search engines, are more controls really what people need? Most searchers just want the right thing to happen. They want it to just work.

See also my earlier posts, "Personalized search at PC Forum" and "Peeking at the future of MSN Search".


Andrew Goodman said...

You're right, of course. It will be at least 2-3 years before users are comfortable using these knobs. I'm still excited by them, though. I think users eventually *will* use them. It at least opens up new dimensions for designers of search products to look at, and may act as an educational tool for the next generations of searchers. I sense change in the wind.

Greg Linden said...

Hi, Andrew. Good to hear from you! Traffick is a great site.

Great point. People may become more comfortable with some of these more complicated tools over time as the general population becomes more computer savvy. A few others made this point as well, including Gary Price.

I do think sliders may work well for discovery. When you don't know exactly what you want and exactly what the options are, sliders help explore the space. I've seen particularly compelling examples for shopping where sliders are used to impose constraints on price, size, weight, or other characteristics to help you focus in on the products that meet your needs.

Unfortunately, web search is a different problem than shopping search. For shopping, the number of products is small (a few thousand) and characteristics of the products are well understood. For web search, the number of documents is huge and the characteristics of full text documents are poorly understood. It is difficult to provide helpful and easy to use sliders for web search.

In the end, I suspect query refinement tools and clustering (e.g. Clusty) will prove more useful to web searchers than sliders.

Of course, just because I think something is a dead end doesn't mean it is. It's great to see so many companies pushing in so many directions right now. I too am thrilled to see so much innovation and exploration in search, including the clever work we are seeing from Yahoo Research.

pat said...

Sliders work pretty well on Kayak. But I agree they don't seem that effective on Mindset.

Craig Danuloff said...

Good post, but I disagree with your 'Grandma' issue on the slider. With nearly everyone in the world searching these days, we can't have search engines only aimed at Grandma. I like the simplicity of the Yahoo approach even though I myself might like (and use) 3-5 different sliders (grandma I ain't). But I think their decision to just have one, and have it offer such clearly different choices is a good one that most users will understand and come to appreciate. Like every change, it will take a long longer than seems reasonable for adoption, but users have to participate in giving the engines more to go on or results will just continue to get worse (and more heavily manipulated). I've posted more thoughts here:

Greg Linden said...

Thanks, Adam, glad you are enjoying Findory!

Data sharing is an interesting problem. Steve Gillmor seems to have been thinking about this a lot with his work on Attention XML. Might take a look at his thoughts if you haven't seen them already.

This is less of an issue than you might think for Findory. Findory is designed to work from sparse and noisy data. You should find it works well even if you have only read a few articles. While more information helps refine the recommendations, it doesn't need to know everything about you to help you discover interesting articles.

Earnest said...

It's true that the slider may fail the grandma test, but so does Google's personlization scheme. With Google there are too many options and it takes too long (more than a couple seconds) to set your personalization criteria. At least Yahoo's slider limits your options to the two things most people use the internet for: commerce and research. It still needs work, but I'd love to have a switch that you could simply turn off and on to filter out commercial content. Ah, whatever happened to Northern Lights-- their use of folders in search was a godsend.

Greg Linden said...

Earnest, have you seen Clusty, the clustering search engine? Definitely worth a look.

beach said...
This comment has been removed by a blog administrator.
beach said...


We found that sliders worked for many different types of users when we originally introduced them with Yahoo! SmartSort... We were surprised by this.
(forgot the colon earlier...)

What makes them work in these cases are the instant results. They maybe confusing at first, but they are recoverable. Once you try them and see how the results change they become more engaging. People really liked playing with them.

I should say that we tested the hell out them. In our own lab and in the midwest with a huge varity of users. All ages. We were happy with the response.



Andrew Goodman said...

I believe in my original blog about Google's sliders (in their personalization lab experiment), I pointed to the example of audio equalizers, an "audiophile" obsession that began to get popular in the 1970's and eventually filtered into the mainstream. I don't know about you, but on my simple digital equalizer in my standard-issue car stereo, I have it set to "bass +3" and "treble 0".

Some stereos were designed with preset "equalizer patterns" that had nicknames. "Boldness," "loudness," "sharp," etc. - so the multiple slider settings were implied and built in, and given a name. People used those buttons too. On my Dad's old Sony stereo, "loudness" always sounded cool with music playing but not so great with people talking. We thought about it a bit and weren't daunted by it. People like to fiddle.

These kinds of things *can* filter out to the general population. Google itself is an example of what sounded like a "niche" search engine that was used only by journalists. (Kind of clever that Google's PR dept. managed to get all the big newspapers and libraries hooked on it, and willing to say so. That gave the next wave of adopters a reason to try it, and then it was able to fend off threats as it fought to get out into the mainstream, which it did.)

A couple of years ago, people didn't know how to attach attachments or file webmail in folders. Heck, they didn't even use webmail. Then one day Mom and sis and uncle and everyone are using Hotmail, Yahoo Mail, etc., for everything.

The grandma test definitely doesn't impress me. :)

Go, now... look in your car, or your dad's car... see if somebody set bass to +3.