I say start with the purpose and work backwards. As the good Dr. Lecter said, "First principles, Clarice. Simplicity. Read Marcus Aurelius. Of each particular thing ask: what is it in itself? What is its nature?"

People really only use data for a few reasons - to be or take action towards pleasure, to reduce risk, save time or make money. Combine any of the following for extra points.

Why then do we love data that is socially generated or a derivative thereof? It turns out to be really good at leading to those outcomes as it communicates experiences others have actually had.

Take Amazon's 5 star rating system. A very nice graphical device that helps me lower the risk I will buy a bad product. You know what is even better? The graph showing the counts of ratings in each star. It tells so much more! I make better decisions. I don't buy anything without first checking if Amazon has a rating.

If you are going to show me a graph - it better lead to one of those outcomes and rather quickly. Once you nail down the purpose then you can decide everything else.

Okay, in general, good.

Now, for the social Internet, what are the purposes? E.g., why do 500+ million people use Facebook? Why do so many people use Twitter?

Okay, once we have thought about those old cases, assuming the huge amounts of data on social graphs in principle available, to what purpose might that data be put?

E.g., Fred Wilson wants to get copies of his social graphs and then manipulate, curate, them himself. Okay, but for what purposes?

Then as in the posts in his thread, how to process the social graph data automatically for Fred's or other purposes, and, again, what purposes?

E.g., maybe the purposes are introductions of one person to another or, say, one person to a group. Again, for what purposes? "May I have the envelope please?" Here is a list -- hobbies, careers, looking for a supplier, customer, employer, employee, date, spouse, cello in a string quartet.

Then if the purposes are mostly just introductions, how are the problem and solution much different than the introductions via the present romantic matchmaking sites?

Uh, Brin indicated that 99% of what can be done has yet to be done, or some such. So, what are the purposes of the other 99%? So, is Brin's 99% just an expansion of the idea of romantic matchmaking sites?

In simplest terms, what are the important candidate purposes for which people would want the results of software that manipulates data on social graphs?

I'm asking because the question seems big, hot, and important, and I don't have any very good answers. Here I may not be alone: Facebook, Twitter, etc. seem important, but finding clear and seemingly accurate statements of the important purposes they serve is not so easy.

But there is potentially a LOT of data so that if we could be clear on purposes, then maybe we could achieve some of them.

Let's do some guessing:

Social search might necessarily be quite close to personal interest search. Then the main purpose of social whatever on the Internet would be some personal interest so that doing a personal interest search would yield arcs in the user's social graph, where each such arc leads to a node good for that user's personal interest and, thus, leads to Wilson's "Building Better Social Graphs".

In particular, we can have bars, art galleries, concerts, adult education classes, country clubs, yacht clubs, high school and college alumni meetings, and, quite particular to the Internet, topical blogs.

So, each of these social activities can further a person's personal interests and build arcs in the person's corresponding social graph.

Then, perhaps, finding such a topical blog, the user could post some comments. Eventually maybe someone would post "For all blog users in and near Seattle, this Friday let's meet for pizza and beer at Joe's place".

So, the purpose would be to meet some people who share some interests and a geographical area.

The Internet has been used for such things. So, the purpose is to meet people with shared interests and geography. So, the purpose seems to be one that could be accomplished better by a slight generalization of Internet romantic matchmaking services.

If so, then how come Facebook and Twitter are big things far from the matchmaking services that are much smaller things?

The social internet garners its masses by feeding, often simultaneously, into the most basic human needs. Our needs for social information exchange are fulfilled, we can find and cultivate our own community. Humans are hardwired to be connected. Socially we evolved to be as such because at its base it is life sustaining. We survive because somebody bit the dust eating a mushroom and somebody else learned from it and then taught us.

The other side of the success of Facebook is that it serves to entertain us. Americans watch on average 34 hours of TV a week. Most of it is nearly worthless and not a value add other than to be amused.

But, people in the course of fulfilling these tasks on FB are leaving behind information trails. We know know what they are talking about, who they are talking to, who they just friended. Now we can compile this information and do something with it - back to the purpose.

Let's take a look at a site like hunch.com - a very good use of collecting data about an individual by answering questions (a form of amusement to many) written by other members. They then leverage it to solve some problems that Amazon's rating system fails at - letting me know what user reviews are more like me. Am I more similar to people who give it 1 star reviews or 5 star reviews. I really enjoy the site and I willingly give it information about me so it can help me more.

This data is usually used as you stated to find similarities . In people, for matchmaking sites, here are things purchased by people who...read the same stuff, watch the same movies, etc. It can turn out well, but for what was known by the Netflix challenge teams as the Napoleon Dynamite problem. A movie that based on similarity clusters was almost impossible to predict how a user would rate it.

Now as you ask what are the candidate purposes for which people want the results of software that leverages social graphs.

Companies are much easier than people because unlike our list of uses and goals earlier, companies just want to make money. Facebook for example.. How do they make money? By selling advertising (other sources comprise only about 10%). How do they make more money? By entertaining more people, for a longer time, and charging more for their ads. If you can serve a targeted ad you can charge more, etc. If you know what feed information to show a user, they will stay on longer, you then serve them more ads. And as such, their purpose for using the social data and derivatives thereof will be for those purposes. Similar to amazon, the purpose of the data are to sell you more goods by recommending things you might buy based on predictions made.

So why does Facebook lack dating as part of their core offering? Because while they could be good at it they up to now have just been a platform company. This has changed a bit with the launch of FB messaging and @facebook email. But look deep into the reason why? Gmail makes Google money because they serve you contextual ads based on your content. Why does Facebook hope you use their mail? Ah yes, you get it now...so you spend more time on their site and they have more data about you so they can serve you more ads. With one small feature change you might say the showed a glimpse of who they are becoming.

Before they were a true social media company – a platform where users generated content to share. Now they want you to spend more time on the site, use it as a common login for other sites, send and receive email through it? Why..again the money. So they can do what they do best. Serve you better ads and have a longer time to do it. Every move you see FB make you will see it is now to broaden the platform and increase your dependence on it so you spend more time on it.

Facebook is becoming the new money printing machine by taking exactly the same path Google originally did – by getting people to the information they want faster. How did Google do it? Google page rank – they figured out that rather than word density – the social element of a link to the content was more important. Google is scared now because the content that is being generated inside of Facebook’s closed walls is massive and it is a powerful predictor of what is relevant.

In both Facebook and Google’s case they make money by helping us perform the transactions we desire by using social data. Social search on the other hand, at least how I see it going, is leveraging all this data to wade through the murky garbage everybody is producing with social media and actually help me find what is relevant. Relevancy is key because what is relevant to me I look at, I take action with, I buy. This makes other people money usually (if it does not then they usually fail to build or go under).

Again the purpose is user specific and the social data helps you determine relevancy. A matchmaking site would use it to find you a date you liked faster, a consumer retail site would use it to help you find the garlic press you are going to like faster.

Finally - back to the graph. What are your thoughts on using them? Personally I would find them useful for discovering trends in myself, my research etc. They help us explore, sort through data. But they are the end product and for most companies just a form of entertaining you with all the data they have about you.

Maybe we're making some progress:

So, we started with social graphs, social media, and social search (below I refer to these three as 'social X') and asked the questions about purposes, the utility of social X, and the means to the purposes.

For entertainment, there is a huge industry from movies and TV. One lesson is, high quality content with broad appeal is expensive to produce. In particular I don't see Internet social X doing much better at generating good entertainment content.

I will insert another point: The more such good entertainment is produced, the more expensive it is to produce even better entertainment. So, we should be moving to about enough high quality digital recordings of such entertainment to satisfy anyone for life and, thus, have less pressing reasons to produce more.

You mentioned:

"This data is usually used as you stated to find similarities . In people, for matchmaking sites, here are things purchased by people who...read the same stuff, watch the same movies, etc. It can turn out well, but for what was known by the Netflix challenge teams as the Napoleon Dynamite problem. A movie that based on similarity clusters was almost impossible to predict how a user would rate it."

I can believe that! Apparently the idea was that persons A and B are 'similar' and person A likes movie X so that then maybe person B will also like movie X. That is not promising because the 'connections' are too crude: E.g., two sisters can be very close yet be very different in the movies, music, clothes, boys, etc. they like.

It appears that the Netflix people fell into a trap: Apparently commonly in the literature of 'collaborative filtering', 'recommendation engines', and 'data mining' has been the suggestion that we should start by putting people into 'clusters' and for each cluster observe what movies, music, etc. were popular in that cluster. Then, given another person, say, a user, find what cluster they were in and then recommend to them what was popular in that cluster. Not a good idea. Interesting to hear that the idea failed!

So, we might conclude that some of the data for such clustering, recommending, etc. can be represented as social graphs, but that step appears to be just a tangential afterthought and not a good example of the utility of social graphs.

Yes, people like to communicate and form communities: Significant examples from the past include back fences, kitchen tables for coffee, lunches, dates, parties, various meetings (e.g., PTA, alumni, school board, political parties, church), school, and work. Such conversations and communities have been significant for romance, building a business, getting a job, education, skills, arts, crafts, politics, and more. But I'm failing to see that the communities that result from Internet social X are very significant.

Net, I don't see current Internet social X doing very well on entertainment, conversations, or communities.

I see the Internet as terrific for a lot, e.g., news, short video clips, technical information, the content on Wikipedia, general information on organizations, information from topical blogs and long tail Web sites, shopping, and more. But so far I don't see social media as leading to very significant utility.

Then in particular I don't see much value in the data on social graphs.

Still, Facebook has 500+ million users and is worth maybe over $50 billion; Twitter, Foursquare, etc. are significant; and Page believes that 99% of social search has yet to be done.

So, here we can look at the weakness in social X, and that might mean that there is an opportunity to do better and have some astounding success. Or, a guess is that current social media efforts are not really addressing the fundamental purposes and opportunities very directly or well so that a better effort could be quite successful.