Hacker News Reader

Comment by cstross | Hacker News Reader

My social graph doesn't tell you anything useful about me.

What it tells you about is people who like the books I write.

Someone sends me a friend request? I auto-friend them unless they're an obvious spambot. It doesn't mean they're my "friend" or someone I know IRL; probably 80% of my FB "friends" live in parts of foreign countries I haven't visited. I don't use FB for social purposes, I use it for marketing communications.

(EDIT: Paradoxically, it turns out that many of my real friends aren't hooked up to me via FB. And none of my relatives -- except one nephew -- even have FB accounts.)

This wouldn't be a problem except for Zuckerberg's sociopathic insistence that we all have just one unitary identity. Groan.

View on HN

The wonderful, terrible thing about this is it literally doesn't matter whether the social graph says anything useful about you. It only matters that it says something which, when added to existing data sources, is a tiny sliver more accurate than the bank's internal risk scoring when aggregated over 100,000 people who happen to include you in them. If you happen to be misjudged by the algorithm, that isn't in itself a problem.

If I were writing a social network layer to a creditworthiness algorithm, the first thing I'd do is create some kind of taxonomy of friend networks. Got 500+ friends? See if the Google results for your name indicate fame (i.e. you show up in title tags on Amazon, IMDB, or the NYT). And bam, reduce the social networks' weighting.

Although I would be willing to bet that a) your fans will have above-average credit scores adjusted for age, and that b) authors whose fans have crappy credit are themselves more likely to have crappy credit.

In the extreme case, if you're a law professor and your FB friends are all attorneys who loved your book, you're probably a good credit. If you wrote a guide to how to fool your parole officer and your FB friends were all fans of your book, your odds of default might be pretty good.

What if you are Adam Sandler or another intelligent, responsible comedian with a base fan base?

No heuristic is perfect, but the question is whether this situation is common enough to swamp the general informational effects. So, three possibilities:

1. Adam Sandler will pay a fractionally higher interest rate. 2. The heuristic will route around this situation. 3. Banks that blindly apply this heuristic will lose money on Sandler because they make a comparatively unattractive offer; the bank that lends to him will get higher market share by being less cautious or more careful in this instance.

Keep in mind that we're talking about an edge case among edge cases. Authors are a tiny minority of social networking users. And authors whose fandom is a contrarian indicator of their creditworthiness are even rarer. Actually, it would more likely work in the other direction: I bet poets published in the New Yorker have very creditworthy friends and still have trouble paying the rent.

Hopefully you don't write self-help books for the unemployed.

"""My social graph doesn't tell you anything useful about me."""

It doesn't matter. Someone can use it as saying all about you, anyway.

Right you are.

Now extrapolate. What if I'm not an exceptional sparkly unicorn?

At least credit scoring on the basis of repayment of prior debts is linked to your own actions.

You're not?! I'm unfriending you on Facebook.

(The upside of this is that your credit rating's on the uptick!)

Reply on news.ycombinator.com