Thread: We Already Have a Social Credit Score

tl;dr: (in the same way we have utility functions - by backsolving)

Unpopular opinion that I actually believe: it’s dumb to make fun of China’s social credit score system since we (the United States) already has one. 1/

Premise:

1. We have different types of “records” that can be checked at different costs.
2. We disallow / allow people to do certain things if certain items are on-record.

Proposition: 1 + 2 means I can probably solve for numbers to back out an actual social credit score(s). 2/

Examples:
- if you committed a felony you can’t get certain jobs
- if your credit score is too low you can’t get certain jobs
- you need a PhD to get certain jobs
- your degree makes certain jobs easier
- you can be an accreddited investor only if you have > X in $$...

3/

My claim is that we can solve some inequalities to get numbers on the system such that, for example, “felony” is something like -500000 points, and if your score is negative you can’t get certain jobs. So there are some hidden numbers associated with crime, degrees, etc..

4/

such that the kind of checks above become just seeing if your “score” is above or under a number; this is *exactly* the process by which we actually compute utilities; we start with preferences and reverse-solve to get utilities that make the preferences correct inequalities.

5/

We accept utilities (and also things like probabilities as measures of belief) as *implicit* things that are “real” because we can solve for them, not because we know them. In that sense, my thesis is that the US in a very real way has these scores already, just implicitly.

6/

Caveat for the mathy: since some preferences are domain-localized, solving for a single number may actually fail, but adding dimensions definitely gives enough freedom (so instead of a single number, you may have a “money score,” a “education score,” a “crime score” etc.)

7/

Side comment: why do people not find “credit score” insane if they find a “social credit score” insane? They’re practically equivalent as ideas!

Hypothesis: it just “feels” bad since China is doing it. If China were the first to make a “credit score” we’d hate that too.

8/

Counterpoint: “we already know this. The problem is China will use it wrong, and a big part of the score will be obeying the Party.”

Countercounterpoint: do you really not think we don’t already have a system where signaling against the “dominant values” loses you points?

9/

Countercountercounterpoint: that’s not the same since we have multiple parties

Counter^4 point: maybe we have 2 “party loyalty scores,” one for the Left and one for the Right. Saying certain things will cancel you (give you -3295.2 points) for the Left, or vice-versa!

10/

At least China is straightforward about it and makes the process efficient. We’re the ones obfuscating it under layers we don’t talk about, pretending we don’t have them already, and waste people’s computational power to compute scores w/ the Court of Public Opinion

END 11/11

Addendum 1: Goodharting applies very strongly here and making an important quantity too clear causes a ton of bad things. (I assumed the on the last tweet was enough to signal this, then I realized when I assume I make an ass out of ume, which isn’t fair to ume)

Addendum 2: I wrote this thread for my stated point. If it were “*should* we do it?” it would’ve been a 30-page blog post of reasons not to:
- Goodharting
- monolithity
- 1-dimensional
would be the meat of it, but these points are often made, so I focused on something different.

Addendum 3: the following jab of me is 100% accurate: https://twitter.com/0k_ultra/status/1280616585309368329?s=21

I am indeed being that guy that is telling you “dinosaurs are birds” because that’s an interesting line of thinking too https://twitter.com/0k_ultra/status/1280616585309368329