Re: Loans

1

I'll be asking all of you who are currently connected for credit reports; failure to respond within 10 days will result in unfriending.


Posted by: SP | Link to this comment | 08-18-15 7:22 AM
horizontal rule
2

The patent linked to by the venturebeat.com article isn't the patent described by the article itself, so I can't tell exactly what's going on here, but people who write about patents on the Internet tend to confuse (1) patents with published patent applications, and (2) specifications with claims.


Posted by: Patton "Tex" Aminer | Link to this comment | 08-18-15 7:31 AM
horizontal rule
3

Patent applications from large technology companies are generally defensive and meaningless, which is not to say that lots of people (including, presumably, facebook) aren't working on using online social relationship data as extra information for creditworthiness judgments.

The positive spin, if you care for one, and with some important caveats on "positive", is that it could allow people who have not yet been able to establish a credit rating to get loans.


Posted by: Sifu Tweety | Link to this comment | 08-18-15 7:32 AM
horizontal rule
4

Loans are a pretty half-assed way of expropriating the expropriators.


Posted by: Moby Hick | Link to this comment | 08-18-15 7:34 AM
horizontal rule
5

Unless you don't pay them back.


Posted by: Moby Hick | Link to this comment | 08-18-15 7:35 AM
horizontal rule
6

It might not pan out as a patent, and it might not be feasible commercially, but given the first two you know they'd do it in a hot minute.


Posted by: Minivet | Link to this comment | 08-18-15 7:39 AM
horizontal rule
7

So, it's a patent on taking an average? Seems legit.


Posted by: Yawnoc | Link to this comment | 08-18-15 7:41 AM
horizontal rule
8

I assume they'd do it solely on the commercial feasibility.


Posted by: heebie-geebie | Link to this comment | 08-18-15 7:41 AM
horizontal rule
9

Wouldn't this be a great product for the deadbeat children of privilege? The kind who are always touching their friends for a few hundred here and there? Or the kind that emotionally manipulate Mom and Dad into financing their next score? Seems like it would be worth it just to get them off everybody's backs.


Posted by: Yawnoc | Link to this comment | 08-18-15 7:47 AM
horizontal rule
10

Let's just call them 'juicers'.


Posted by: Moby Hick | Link to this comment | 08-18-15 7:52 AM
horizontal rule
11

The patent linked to by the venturebeat.com article isn't the patent described by the article itself,

Indeed. The lender bit isn't mentioned at all. The actual patent appears to be this one.

It is kind of weird, in that you can see the legitimate purposes of the other embodiments, and more importantly they're based on positive matches within the social network to the benefit of the user - basically, if there's a match, you're in, while stuff from unknown parties is blocked. The lending one is totally different in character, and can only be detrimental to the user (assuming it's in their interest to be given credit). And that's even before we get to the ethics/legality of basing credit decisions on who you know rather than who you are.


Posted by: | Link to this comment | 08-18-15 7:56 AM
horizontal rule
12

Historical perspective: Making lending decisions based on knowing the customer's social network was the entire theory of community banks and of credit unions: You could get a loan if the banker knew you from the church social, or the country club. It was also the basis for niche banks serving various ethnic communities, including African-American. The muitinational banks have been trying to compete in this kind of information gathering for decades now with different kidns of proxies. Facebook social networks seem as likely to provide useful information as anything else, but still not very likely.


Posted by: unimaginative | Link to this comment | 08-18-15 8:44 AM
horizontal rule
13

So this is a proposed supplement/replacement for FICO scores?

I honestly don't see the problem with that. There are foreign microlenders that explicitly lend to groups of people with no capital-- Grameen for instance.

The CFPB wiki page is missing some pieces. But the current house hasn't completely disabled the agency, and they're shutting down some scams.


Posted by: lw | Link to this comment | 08-18-15 8:44 AM
horizontal rule
14

The problem would be if it's discriminatory instead of faciliatory.


Posted by: heebie-geebie | Link to this comment | 08-18-15 8:45 AM
horizontal rule
15

Yes, it's very much pitched as a way of denying people loans who would otherwise get them, not granting loans to people who otherwise wouldn't:

When an individual applies for a loan, the lender examines the credit ratings of members of the individual's social network who are connected to the individual through authorized nodes. If the average credit rating of these members is at least a minimum credit score, the lender continues to process the loan application. Otherwise, the loan application is rejected.

Or, in more detail:

FIG. 9 is a flow diagram that illustrates the steps carried out in authenticating B for access to a loan. In Step 910, the lender receives a request for a loan from B. The request includes certain identifying information of B, such as B's e-mail address. In Step 920, in accordance with the methods described in the application, "Method of Sharing Social Network Information with Existing User Databases," (U.S. patent application Ser. No. 10/854,610, issued as U.S. Pat. No. 8,478,078), filed Jun. 14, 2004, this lender makes a request to a social network database for a graph representation of B's social network and receives the graph representation of B's social network. In Step 930, a black list that is maintained for B is requested and received from the social network database in the same manner as in Step 920. In Step 940, a gray list is derived from the black list and B's social network In Step 950, a breadth first search (or alternatively, a depth first search) is conducted on B's social network to generate a white list. All members of B's social network who are connected to B along a path that does not traverse through any unauthorized nodes (i.e., individuals identified in the gray list) get included on this white list. Optionally, the lender may specify a maximum degree of separation value (e.g., N.sub.max). If it is specified, the white list will include only those members of B's social network who are within N.sub.max degrees of separation from B. In Step 960, the credit ratings of individuals in the white list are retrieved and weighting factors are applied to the credit ratings based on the degree of separation between the individual and B. As an example, a weighting factor of 1/10.sup.N may be applied to the credit ratings, where N is the degree of separation between the individual and B. If the average credit rating is above a minimum score, B is authenticated and the processing of B's loan application is permitted to proceed (Steps 970 and 980). If not, B is not authenticated, and B's loan application is rejected (Steps 970 and 990).


Posted by: Ginger Yellow | Link to this comment | 08-18-15 8:57 AM
horizontal rule
16

Maybe that was unclear-- I don't see control of the information used to make lending decisions as an especially critical frontier of consumer finance. More important to prevent deceptive marketing and predatory lending thresholds (balloon loans to people without much income but some capital built up in their houses, say).

Separately, as others have noted, the linked paragraph does not actually appear in the patent. The patent app doesn't have any notes indicating revision. I think bad internet reporting, but since the linked paragraph appears on many pages, hard to trace source. The link from Mark Sullivan's bylined article is to a USPTO page that includes search query(company and date range, no imagination or thought there)-- so it looks like the original writer just browses patents and writes off the cuff. The "publisher," VentureBeat, has been operating since 2006, so I guess this level of work pays somebody's bills. Bummer, fails cursory refereeing.


Posted by: lw | Link to this comment | 08-18-15 9:02 AM
horizontal rule
17

lw, see the link in 11 for the correct patent.


Posted by: Ginger Yellow | Link to this comment | 08-18-15 9:05 AM
horizontal rule
18

Oh, I see, patent 8782753 is mentioned in the article without being cited.

OK, not fabricated, better.

I still don't see the problem socially. I see the practice of redlining as basically political rather than technical, and I see the predatory lending in the recent housing crash as being a consequence of weak regulation, not runaway technical innovation.

There's such a thing as too much credit definitely, how much is right is a complicated question. I just don't see this particular technical innovation as being close to a practical boundary.


Posted by: lw | Link to this comment | 08-18-15 9:21 AM
horizontal rule
19

I still don't see the problem socially. I see the practice of redlining as basically political rather than technical, and I see the predatory lending in the recent housing crash as being a consequence of weak regulation, not runaway technical innovation.

I don't disagree with this, but can you not see how technical innovation, and this particular methodology in particular, could systematically reinforce existing racial and class disparities? Even as another factor in the underwriting process it might cause problems, but as a hurdle to even get in the door it seems extremely problematic.


Posted by: Ginger Yellow | Link to this comment | 08-18-15 9:46 AM
horizontal rule
20

15: I don't see how it's necessarily a way of denying people loans who would otherwise get them. The entire underwriting process is about kicking people out -- so while changing the underwriting process means changing how to kick people out, that's not equivalent to kicking *more* people out. In fact it's entirely likely that the Facebook data is most relevant when people are already unlikely to get a loan.


Posted by: Disingenuous Bastard | Link to this comment | 08-18-15 10:45 AM
horizontal rule
21

19. I see that this is how fb's patent app describes the tool, and if that were implemented, yes, troublesome. But a sand grain of troublesome compared to current practice of actual shitty lenders (the bust I linked above: bought the domain name "fafsa.com", fafsa is an acronym for a student loan application form.)

Basically, I see fair credit for people with few resources and unclear records as a very important, and technically a difficult problem for hypothetically existing well-intentioned banks to solve. This particular "innovation" from some fb marketer doesn't interest me. CFPB busts and nontraditional lenders (LendingClub to crowdfunders) seem to me like more interesting spots on the boundary around fair economic opportunity.


Posted by: lw | Link to this comment | 08-18-15 11:11 AM
horizontal rule
22

Patent specifications are shockingly meaningless and have extremely little to do with existing or planned technology, as some people have said above.

Personally I feel like the current credit score system is at some worst possible level of being simultaneously privacy-invasive, opaque, and erratic and just plain wrong, so it's hard not to see more information as an improvement.


Posted by: Roberto Tigre | Link to this comment | 08-18-15 11:13 AM
horizontal rule
23

When I saw the article I was surprised that they were getting a patent on something so well known. There were articles several years ago about how loan servicing companies were denying mortgages based on people living close to people with bad credit scores and how they could be disqualified by making certain types of purchases with their credit cards.

But the claims of the patent are very different and this scenario is merely a use case described in the third-to-last paragraph of the patent. The claims are about how you can't get information from one of your connections in a social network if the information comes from an "unauthorized node."

I'm a patent attorney and 80% of my job is obtaining patents in the social-network space for a very large company that is not FB.


Posted by: LizSpigot | Link to this comment | 08-18-15 11:14 AM
horizontal rule
24

Let's all think of shitty things people could do with social network data, patent them, and then block all uses. Until the patents expire, I guess.


Posted by: fake accent | Link to this comment | 08-18-15 12:16 PM
horizontal rule
25

Don't patents cost money?


Posted by: Moby Hick | Link to this comment | 08-18-15 12:18 PM
horizontal rule
26

Freedom ain't free, Moby.


Posted by: fake accent | Link to this comment | 08-18-15 12:38 PM
horizontal rule
27

23. Is there prior art for going into bars with good light at happy hour and then complimenting age-appropriate women who make eye copntact on well-chosen footwear ?


Posted by: lw | Link to this comment | 08-18-15 1:51 PM
horizontal rule
28

27: I got you bar, light, happy hour, woman, and age, but no complimenting footwear.


Posted by: LizSpigot | Link to this comment | 08-18-15 2:14 PM
horizontal rule
29

So how did patentese get that way?


Posted by: nosflow | Link to this comment | 08-18-15 2:19 PM
horizontal rule
30

29 -- I'm not a patent guy, but it's an old system, in pretty modern form in the US since the 1830s, and has evolved more or less like all common-law legalese, with lawyers finding useful phrases and borrowing them, and then courts writing opinions in such a way so that some of these phrases become magic words with great importance, on and on ad nauseum.

Patent language is so weird mostly because it's extraordinarily hard to describe what an invention "is" in English language words. But that's the essence of a patent -- it's the specific "claim" in the invention, which is written as a sentence, that under our system defines the right and gives you rights against the world. So drafting those weird words becomes incredibly important.


Posted by: Roberto Tigre | Link to this comment | 08-18-15 2:54 PM
horizontal rule
31

Patents also (in my field, anyway) are written to balance gaining legal protection and disclosing trade secrets. The technical bits are weirded by lawyers, as far as I can tell.


Posted by: ydnew | Link to this comment | 08-18-15 3:11 PM
horizontal rule
32

Some patent attorneys like to hide the ball because then they can get patents that are broader. Specifically, because inventors can be their own lexicographers, they can use new words. Lazy examiners can't find a perfect match for those words, and as a result, they allow broader patents.

However, this is not my practice because I write patents that could be understood by a jury. If you're using a patent in court, a broad patent is pretty useless if a jury doesn't understand what you're talking about.

My patents have sections with all the legalese thrown in, but the main text is very understandable, especially when you read it while looking at the figures.


Posted by: LizSpigot | Link to this comment | 08-18-15 3:26 PM
horizontal rule
33

Patents also (in my field, anyway) are written to balance gaining legal protection and disclosing trade secrets undermine the reason society grants patents to begin with.

Seriously, the whole point is to incentivize the eventual release of trade secrets, no?


Posted by: Spike | Link to this comment | 08-18-15 4:03 PM
horizontal rule
34

If I said your patent had a beautiful figure would you let me infringe it?


Posted by: fake accent | Link to this comment | 08-18-15 4:11 PM
horizontal rule
35

I take back Michelle Obama endorsement because having fake accent in debates would be even more fun.


Posted by: Thorn | Link to this comment | 08-18-15 6:24 PM
horizontal rule
36

Unmarried candidates are unelectable. The best I could do would be to mimic Trump from the left and say a bunch of outrageous things.


Posted by: fake accent | Link to this comment | 08-18-15 7:11 PM
horizontal rule