In mid-2009, I had a wonderful idea of integrating people's evaluations into search engines. That was based on my own search habits. Most of my search queries are about science and technology. The quality of these kinds of materials tends to be stable, which means a good article will always be good regardless of the time of search. That's how we define classic articles. Queries about science and technology are usually topic-centric. When we want to know about a topic, the best answer is often the classic articles.

But to figure out classic articles requires expert knowledge. An expert's evaluation about his/her speciality is much more helpful than a layman's. Thus your search experience will be greatly improved with the help of experts' opinion. However, the problem is that you have no way to see their evaluations. You may not know them personally. Even if you happen to read their evaluations somewhere on the Internet, it's very likely that these opinions are in the form of non-structural articles, which means they're hard to be integrated into search engines and it's still difficult to automate the utilization of these evaluations.

So it will be of great help if experts leave their opinions in a structural way. How can this be done? It's simple. They just need to mark a webpage as good or bad in the process of reading. Now you see why Google+ emerges? I've never believed Google+ is simply meant to be a social network from the day of its birth. I strongly believe the main reason of the emergence of Google+ is to improve the quality of Google Web Search.

In my original design, a social network benefits a search engine in a free manner. What do I mean by free? Consider what will happen if some expert doesn't what to share his/her opinion. Does that matter? Usually not, because there will be other experts in the same field who are willing to share and you just need to follow them instead. Another problem is that the opinion of an expert in area A may not be very helpful to a query in area B (they even act as noise). But you can put experts in different groups and set which group to take effect, possibly in different weights, for different queries.

The main advantage of this design is, you have the choice of who affects you. People who post helpful opinions will be followed by more users and become welcome and gain fame. People who post useless opinions, and thus play the role of unwelcome noise, will be discarded by users and only cause limited harm. Therefore the speakers have motivation to make good evaluations. This is the power of natural selection.

At the same time of listeners selecting speakers, speakers also select listeners. Listeners who do bad intentionally by following unwelcome speakers to increase their popularity will be punished, because they'd have much noise in their personal search. This discourages them from doing so.

Therefore, both speakers and listeners have good motivation guiding their behaviors. Thanks to this loose-coupling design, the whole ecosystem will evolve rather than degenerate.

And what's even more important, the improvement of search quality will attract people to join the social network! This is the power of a service-wide win-win strategy. If there is mutual promotion between two services, users of one service will be encouraged to use the other, and vice versa.

Maybe that's why Google introduces Search, plus Your World. Basically it adds three kinds of information in the search results: personal results, profiles in search, and people and pages.

However, this design doesn't follow my original idea faithfully. For example, you cannot filter evaluations from followees by the groups they are in. And there is too much noise in the personal search results. When I want to find the explanation of a scientific term, the personal results usually contain my followees' own experience with it rather than well-accepted high-quality resources related to it. So I think it pays more attentation to personalization rather than personal recommendation. It's more like a brochure of my followees' posts rather than a high-end recommendation system. Of course there are notices like 'XXX shared this' below some search results, but you know their existence only when you see them passively.

My prediction for the way we search is that it will advance through three stages:

• Stage 1. A few people will mark webpages as good or not. But their opinions cover only a small part of everyday search, which has limited benefits.

  In this stage, the usual (non-personalized) search will play the main role.

• Stage 2. A majority of people begin to realize the benefits of personalized search. Due to the advantages of the above design, natural selection will begin to take affect for both speakers and listeners. In the end, mature and stable speaker communities will be formed in various fields.

  In this stage, both non-personalized and personalized search will play the main role.

• Stage 3. Marked webpages will be more and more important. Almost all search results will fall into two categories: marked and real-time. Real-time results will gradually become marked. But it may be a big challenge to maintain the relationship between speakers and listeners.

  In this stage, personalized search will play the main role, except for real-time results.

Let's see whether this will happen in the next few years.