Wednesday, November 15, 2006

Two Social Search Ideas

I'm on a short vacation in the Mediterranean, and as always during trips, I get a bunch of ideas that seem smart when you're chilling on a beach but may be useless when confronted with the real world.

Here are two simple ideas that could make search more useful by accessing your social networks. Both ideas seem pretty obvious. So obvious and trivial, in fact, that I'm pretty sure someone has already tried this. I figured I'd still put them up for discussion.

Click Popularity

A recent article in the Economist points out that people like to follow the herd when confronted with many options. They buy the most popular cereal in a supermarket, and download the most popular songs in an online music store.

What if we extended this concept to search results? If search engines showed click counts for each item on the results page, SEOs would instantly start clicking away, making that measure completely useless.

But what if we integrated search and social networking? We could show just the click counts of your friends. Your friends have little incentive to skew results. They will have similar tastes and preferences as you do, so they will search for similar things and likely click on the same items on a results page. And you could be sure you clicked the "right" result – i.e. the one your friends clicked.


Query Trail Sharing

Search results are seldom perfect on the first try: Even Google can't read your mind. When searching for something specific, users often spend considerable amounts of time refining their queries.

For example, I was recently looking for the name of the Python function that lets me get a class member given a string with its name. The function is called 'getattr', but that had somehow escaped me. Here is the query trail for that day, reconstructed from my Google search history (The first and last queries are unrelated):By looking at word overlaps and the timestamps, one could now find out that the inner three queries belong together. When a friend searches for the same item, one could now show related queries.


Problems

Implementing both features is fairly straightforward and could likely be done with a bunch of Greasemonkey scripts. But the two huge problems are privacy and the number of friends needed to make this useful.

I doubt that users would dare to use this if they thought that their searches are watched by friends. Therefore, click counts and trail sharing should be anonymous: You don't know which one of your friends clicked where. Plus, it may be useful to filter those Jenna Jameson-related queries.

Second, you don't want to be the only one signed up for this service: You only profit from the feature if you have lots of friends signed up for it as well. Sure, we could also look at data from friends-of-friends and further layers, but that increases spamming opportunities and decreases privacy. Maybe it would make sense to integrate this with existing social networks, such as Xing or LinkedIn, and have people download a browser plugin. If Google or Yahoo did this in our post-AOL-leak world, there could be an public outcry.

Let me know in case you know a product that already does this.

3 comments:

Ed said...

Your first idea is exactly why del.icio.us is so powerful.

I have no problem finding over 1,000,000 results for 'plumber in london', what I want is the one my friend recommends, but I want it without having to bother all my friends and ask them. Solution, I search on del.icio.us

Reinhard said...

Yahoo's MyWeb does something similar, however, your friends need to save their favourite sites in their MyWeb, so it is not really based on simple click-throughs.

r.
(wondering why blogger's captcha does not get displayed in my FF, must be my cookie ban)

matt said...

Gabor, Lots of entries to catch up on. Have you seen root.net and the attention tracker? We know Seth Goldstein, the founder, and his idea walks a similar line as your first idea. check it out.