Wednesday, March 25, 2009

Facebook Thoughts: Middle School and Post-Grad Editions

Middle School

So, I told Friend A that I didn't want to discuss Friend B ever again. Which made Friend A thinks that Friend B is completely out of the picture. But I still see Friend B a lot. Both Friend A and Friend B are Facebook friends, but not really social friends other than through me. Recently, Friend B posted a photograph of me on Facebook.

One of the perils of Facebook is that it's hard to tell what Friend A sees: does he see that Friend B posted a picture of me? And I suppose that means I need to explain what's going on. Yes, this is entirely middle school of me.


I notice that my online real estate agent has a way I can post houses that I'm interested in on Facebook. Just because you can do something online doesn't mean you should. And, besides, once the offer is accepted, the listing is no longer available online. Which is exactly the time you want to share it on Facebook.

Maybe I'll have a chance to try it out tomorrow afternoon.

Friday, March 20, 2009

100% Hip-Hop? And what does Linked-In Know?

I logged into Orkut for the first time in years. Possibly decades. Like any good social network, it suggested a few people it thought I might know. Most of them were pretty obvious choices. Bietz, Hand, Casner, Gross. All predictable and unsurprising. Then, nestled in the middle of the list is a "Camila 100% Hip Hop." I'm wondering if she's like MySpace's Tom. Or just a weird freak of the algorithm.

Also, I looked at friend recommendations in Linked In a few days ago. There was one that I actually knew. But I can't figure out how Linked In knew I knew him. There is no common friend. There's not even a two-hop friend. He lives in a distant country, and I've actually only seen him twice in my life. I haven't told Linked In about my gmail account. (Although he may have...) One of those, "huh?" moments.

Yeah. I'm spending too much time thinking about social network sites recently. I blame the US Government.

Thursday, March 19, 2009

Morally Dubious Business Plans

I'm not convinced that this is totally bankrupt. I'm also not convinced that it hasn't already been done by a three-letter agency (CIA, NSA and DHS are the most likely ones).

I've been reading a lot of Broad Agency Announcements (BAAs) from groups like DARPA, IARPA, and NGA which are asking for ideas on how to build a "human terrain map." In other words, what are people thinking, how are ideas and influence spreading, and how would this impact either Phase 3 (major operations) or Phase 4 (stability & reconstruction) operations.

I've been thinking that systems like Facebook and Twitter offer an interesting way to gather a lot of data on the pulse of a community (even a nation) without a lot of effort. One of the problems with blogs is that a lot of people start blogs, but rarely stick with them. Like me. On the other hand, microblogging is a lot easier to keep up with, especially since there is some social pressure to regularly update your status.

What we do is set up a cell phone company that offers reduced rates for regularly checking in on a microblogging service. We'll also push out all of your friends updates to you ABSOLUTELY FREE! And we'll manage your social network for you. And we'll have fun geographic location services as well. Targetting the critical 15-30 demographic, we'll be the network operator of choice. Since we're not trying to make a huge fortune here (remember, we have .13% of the TARP covering costs here), we'll be the first choice, since we're cheap.

There are a few interesting challenges though. First, how do you funnel the data back to your three-letter organization? I think the best way to do this is to create a shell company that does most of the work and advertising. It would be owned / funded / operated by locals. They may not even have to know what was going on: the initial idea would come from a US cellular magnate (Craig McCaw, I'm looking at you.) Then, the actual operations would be subcontracted to a friendly US company with experience in network operations. (Alcatel-Lucent, Cisco, or IBM would be obvious choices. I know, from first hand experience, one of them has extensive experience in modifying telecom hardware to comply with intelligence agency requirements.) The actual data feed back to the US would be done by off-site replication to another data center, and a network tap would be installed there. That's the morally dubious part.

The other interesting problem is how much useful information can you extract from 140 character updates? It's an interesting problem. Most of the text mining systems seem to work better with big corpuses. Also, we're going to try to impute valence and meaning to short messages, many of which may be sarcastic, ambiguous, and deliberately vague. (And, of course, given that we are targeting the "Twilight" demographic, will be 99.999% about relationships, and .001% about hockey, there's not going to be a lot of actionable intelligence in the first place.)

Honestly, I wouldn't be surprised if this is already happening. It seems surprisingly straight forward, and not all that expensive. I mean, compare it to Project Jennifer. Given the right set of data, you could probably even make it break even by selling marketing data gathered. Hell, it might even be profitable in its own right.