IMF researchers: digital footprint yields better credit assessment • Hacker News Board

Re: IMF researchers: digital footprint yields better credit assessment

#91

Post by 8note » Sat, Feb 13, 2021, 6:01 AM UTC

WanderPanda wrote: post #75

dTal wrote: post #53

Earlier quoted context omitted.

I think the key is "feedback loops". We're not failing to price in inherent differences, out of some moral ideal - we're trying to fix a larger inefficiency, which is that some groups of people are unable to reach their full potential. It's not a cost - it's an investment . It's regulation to fix a negative externality. [digression into risky territory, here be dragons] ...at least, that's the idea if you believe tha…

Doesn't need to be inherent to "races". Why wouldn't culture / way of living be the elephant in the room?

"races" is a cultural concept designed to denigrate certain people.

Using race captures an obvious group of people by how they've already been discriminated against

Re: IMF researchers: digital footprint yields better credit assessment

#92

Post by naresh_xai » Sat, Feb 13, 2021, 6:03 AM UTC

hntrader wrote: post #46

aneemzic wrote: post #40

Earlier quoted context omitted.

No. Credit worthiness should be because of an individuals past actions and not some attribute they may have been born with. Why should a tall individual who's never missed a payment have some invisible penalization applied because some other tall people are worse at re-paying loans. Don't you see the issue here? You're penalizing individuals not based on their own measurable behavioral signals, but simply due to some…

We must disambiguate between political objectives and the practice of credit risk modelling. Credit risk, f(X), is an unknown population function that needs to be estimated using observed data X. If including tallness into X improves our estimate of f(X), then we've gotten a better model. You've asserted that X should only contain an individual's past actions instead of their inherent traits such as tallness. This ma…

Unless you have causal proof, its irresponsible for a business to use such factors in modeling outcomes.

Re: IMF researchers: digital footprint yields better credit assessment

#93

Post by throwaway189262 » Sat, Feb 13, 2021, 6:03 AM UTC

thegrimmest wrote: post #44

pmichaud wrote: post #25

Earlier quoted context omitted.

When I have this conversation with myself, here's what I come up with: Shouldn't we be allowed to everything into account when determining credit worthiness? Even if it's socially uncomfortable? We get locally more accurate pricing / risk assessment that way, but we also create feedback loops that we as a society have decided we want to not contribute to. The cost for this is the pricing inaccuracy which we collectiv…

Well, I personally haven't consented to socializing the cost of anything. I'm not sure what the ethical grounds are for imposing this on me, and in so doing dilluting available accuracy of data about people. These feedback loops you're referring to, what order are you proposing for your chicken and your egg? What's causing what in your view?

> Well, I personally haven't consented to socializing the cost of anything.

This is a strange take. If you had to "consent" to laws there would be utter chaos. The places where you can pick and choose which laws apply to you have no real laws. And all of those places are dangerous and destitute. That's why laws are decided by the majority, there will always be dissenters for every law out there. Some harmless, but others that want to pillage, burn things down, and worse.

There will always be laws that you don't like, unless there's no laws at all.

Re: IMF researchers: digital footprint yields better credit assessment

#94

Post by beambot » Sat, Feb 13, 2021, 6:19 AM UTC

hn_throwaway_99 wrote: post #15

I hope that regulatory agencies start cracking down on these "alternative sources" for credit-worthiness data. Researchers have shown how it's very easy to attribute things like political party, race, sexuality, etc., things that are often illegal to take into consideration when determining credit, to other things like where you live, TV shows you like, apps you have on your phone, etc. And strangely (or not) this bl…

I was once peripherally involved in a conversation about ML algorithms being used to determine credit worthiness. One of the bankers explained that race as a factor is illegal, but the algo partially reduced down to a handful of heuristics -- eg that "last names ending in 'z' were less credit worthy" -- and that while these all roughly correlated to race, they were perfectly permissible since the algo was a black box & no one ever ran the correlations (conveniently). TLDR: sickening.

Re: IMF researchers: digital footprint yields better credit assessment

#95

Post by visarga » Sat, Feb 13, 2021, 6:26 AM UTC

> the type of browser and hardware used to access the internet, the history of online searches and purchases ... once powered by artificial intelligence and machine learning, these alternative data sources are often superior than traditional credit assessment methods, and can advance financial inclusion, by, for example, enabling more credit to informal workers and households and firms in rural areas

Here is one of the most sensitive applications of ML. This one, like face recognition, can directly harm people.

These guys only seem to think about the "superiority" of their method in minimizing lending risk. But the other party into this deal can also have something to lose.

What's the difference between this and the Chinese social scoring system? They want to use private search logs. "Your privacy for a credit score?"

Maybe the model is biased, how do we know? Is there an independent bias assessment process? Is there a way to sue for unfair rating? What are the values of the people working on the training set? Is there a "speak to human" step in the process where you can plead your case? I don't want an automated social discrimination machine.

Re: IMF researchers: digital footprint yields better credit assessment

#96

Post by naresh_xai » Sat, Feb 13, 2021, 6:27 AM UTC

chii wrote: post #85

aneemzic wrote: post #71

Earlier quoted context omitted.

> This only lasts until someone figures out how they are different and breaks the association right? Would you be OK if banks started charging you 4% more interest because say you have irish heritage? Your behaviour hasn't changed and nobody can explain why it matters that your father was irish, but the models determined it's statistically significant, What if at some point 30 years down the line it's determined it w…

But you're only looking at this from a single perspective, where every single bank starts charging the same 4% simply because of irish heritage. What if, one bank started doing it because they used this model. But another bank doesn't, because they don't believe this model? If you were an irish person, you would just move banks. And if it turns out, indeed, that the first bank was over-charging the 4% (ie., no irish…

The number of loans a bank gives out to such a specific factor might be low enough that it would take 30 years for a bank to have an actual study. That's 1-2 generations of abuse.

Most banks just copy each other.

Re: IMF researchers: digital footprint yields better credit assessment

#97

Post by Animats » Sat, Feb 13, 2021, 6:34 AM UTC

Would information about CEO behavior be useful to investors? That's worth considering. Browsing history? Phone call patterns?

It's being worked on at the Harvard Business School.[1]

"We measure the behavior of 1,114 CEOs in six countries parsing granular CEO diary data through an unsupervised machine learning algorithm. The algorithm uncovers two distinct behavioral types: “leaders” and “managers”. Leaders focus on multi-function, high-level meetings,while managers focus on one-to-one meetings with core functions. Firms with leader CEOs are on average more productive, and this difference arises only after the CEO is hired. The data is consistent with horizontal differentiation of CEO behavioral types, and firm-CEO matching frictions. We estimate that 17% of sample CEOs are mismatched, and that mismatches areas associated with significant productivity losses."

[1] https://www.hbs.edu/ris/Publication%20Files/17-083_b62a7d71-...

Re: IMF researchers: digital footprint yields better credit assessment

#98

Post by visarga » Sat, Feb 13, 2021, 6:43 AM UTC

beambot wrote: post #94

hn_throwaway_99 wrote: post #15

I hope that regulatory agencies start cracking down on these "alternative sources" for credit-worthiness data. Researchers have shown how it's very easy to attribute things like political party, race, sexuality, etc., things that are often illegal to take into consideration when determining credit, to other things like where you live, TV shows you like, apps you have on your phone, etc. And strangely (or not) this bl…

I was once peripherally involved in a conversation about ML algorithms being used to determine credit worthiness. One of the bankers explained that race as a factor is illegal, but the algo partially reduced down to a handful of heuristics -- eg that "last names ending in 'z' were less credit worthy" -- and that while these all roughly correlated to race, they were perfectly permissible since the algo was a black box…

Here's a case where the Google AI ethics researchers should have unleashed their hounds, not on Yann LeCun and large language models (stochastic parrots? lol). Design evaluation datasets and checklists, do research into "bias washing", call them out on biased models in deployment, etc.

Re: IMF researchers: digital footprint yields better credit assessment

#99

Post by dredmorbius » Sat, Feb 13, 2021, 6:45 AM UTC

JohnJamesRambo wrote: post #86

dredmorbius wrote: post #70

Earlier quoted context omitted.

How are you planning on allocating those bitcoin distributions? Do you have the context for the Finney quote? Apparently from a set of released emails: https://news.bitcoin.com/researcher-publishes-never-before-s... The larger context, from a 1992 Cypherpunks email: "Here we are faced with the problems of loss of privacy, creeping computerization, massive databases, more centralization - and [David] Chaum offers a co…

The full quote supports my assertion even more. It’s a beautiful quote and extremely relevant today, even more than when he said it.

I don't think I can agree. Though I might have once.

There's a fundamental misunderstanding of power, its dynamics, and technology's relationship to it.

Re: IMF researchers: digital footprint yields better credit assessment

#100

Post by NeuNeurosis » Sat, Feb 13, 2021, 6:49 AM UTC

WanderPanda wrote: post #78

slg wrote: post #33

Earlier quoted context omitted.

That is a good summary. An analogy I might add that could further clarify things would be preexisting condition protections for health insurance. Preventing insurance companies from turning down people who they know will require expensive medical care is going to lead to worse pricing for the rest of us. However much of society has decided that is a cost worth enduring in the name of fairness.

Problem with this is, this cuts open the feedback loop of being responsible for ones own health. Credit assignment problem all the way

Are you American? Because high cost of health insurance does not deter risky behaviors towards health or a glaring lack of responsibility.