Facebook always seems to be tweaking its users' experience. Over the past couple of years, the site has changed a lot, with designs geared towards sharing content without also encouraging spam and more real-time features and low-effort ways to interact with friends.

But there are basic parts of the social graph that could use more attention. I was reminded of this by a talk by LiliCheng, general manager of Microsoft's Future Social Experiences Labs, last week at Defrag2009, a technology conference in Denver. Cheng went through some of the intelligent features that are part of Wallop, a social network spun out of Microsoft Research. (Wallop as originally envisioned no longer exists; instead, the company now makes applications for other social networks.)

Wallop was designed to automatically cluster people into groups, Cheng said, making it easier to communicate with relevant friends. It also took emphasis off people who weren't active on the site, displaying grayed-out nodes to show when a user hadn't posted for a long time. "These are really simple things that I think are still missing in a lot of our social tools," Cheng said.

Other social networks could perhaps be much more useful if they could group and manage contacts automatically. It would be good to know, for instance, if a friend rarely logs into his Facebook account, so I could try sending a message a different way. Although Facebook and other social networks do include features that allow users to make and manage lists of friends, few users take advantage. The user experience would perhaps benefit a great deal if these features could become more automatic.

Following a redesign of the Facebook home page earlier this year, product manager Peter Deng wrote: "We built in some default filters based around location, people you connect with most often, and your existing Friend Lists."

Last week I wrote about how data mining could help people manage e-mail. The same is true of other tools that manage lots of information. These techniques can have huge benefits, but it will be most helpful if users can see how they work and tweak them a little.

Everyone knows that online conversations happen fast, but Ilya Grigorik, CTO and founder of PostRank, a company that tracks online conversations around pieces of content, shared some interesting concrete numbers this afternoon at Defrag 2009, a technology conference taking place in Denver. Grigorik randomly selected 100,000 posts that the company had tracked and calculated when conversation happened around them.

It's no surprise that 80 percent of engagement around a post happens on day one, and that 60 percent of that happens within the first hour. What was surprising, however, is that this is actually a decrease from the numbers Grigorik has for 2007. According to his data from two years ago, 95 percent of engagement happened on the first day, and 90 percent of that was within the first hour.

These numbers seem strange considering that the Web appears to be operating at a faster pace. Grigorik's numbers show, for example, that about on average 66 percent of the conversation around a post happens on "chatter" channels such as Twitter, which is nearly the opposite of the trend two years ago, when most conversation happened on the site where a post was published.

Grigorik said he thinks the explanation lies in the effect of the strength of weak ties. He believes that online conversation has become so distributed that it takes time for information to filter out to every social group that's going to talk about it. If he's right, it's a ray of hope for the real-time Web.

On the surface, it might appear that more real-time streams will lead to a stream of data that appears and disappears, leaving no time to ponder the meaning of any of it. If Grigorik is right, however, real-time streams and the social infrastructure around them may help information find its way to more people who would be interested in discussing it.

Google's stepping up its social networking efforts with new features for Friend Connect today, and the features provide some clues as to how Google thinks social data can be used to make money.

Friend Connect provides a way for website owners to give their site social features without having to build an entire social network from scratch. This type of add-on social network tool has been increasingly popular in the last year and a half--Friend Connect competes with offerings such as Facebook Connect, which was announced around the same time, in the first half of 2008. Google says that about 9 billion web sites use Friend Connect, and that the service receives about half a billion unique page views each month.

The new Friend Connect features collect more data about a site's visitors and provide several ways to use it. A polling gadget gathers information about visitors' interests, which is then shared between sites.

Using this feature, a music site could find out which bands are their viewers' favorites, and a fashion site could discover a user's favorite clothing brands. A new direct messaging feature also allows Friend Connect users to contact others with similar interests. The music site could, for example, send newsletters targeted to users who've expressed an interest in certain 90s grunge bands. Or visitors might be served links to the most recent articles about these bands.

But perhaps most importantly for advertising dollars--and one must always remember that Google is an advertising company at heart--user profiles come with an integrated set of tools that a site owner can use to provide personalized information, ads, and services.

Most conveniently, Google has now integrated AdSense with FriendConnect, allowing site owners to fine-tune the ads displayed based on users' interests, as well as site content.

Google's vision of advertising has always been about presenting ad content at the moment people are actively seeking such information, and the company has always employed sophisticated analytics to do this.

FriendConnect's new features look like a solid step toward monetizing social data. While social networking sites still struggle with this--users of those sites are usually looking to socialize, and not to buy things--FriendConnect's advantage is that the social data can be used to catch users when they're looking for useful information or even thinking about making a purchase.

The video below demonstrates the new features.