Math is racist and driving inequality..

Search
Go

Discussion Topic

Return to Forum List
This thread has been locked
Messages 21 - 40 of total 67 in this topic << First  |  < Previous  |  Show All  |  Next >  |  Last >>
Ed Hartouni

Trad climber
Livermore, CA
Sep 8, 2016 - 07:29pm PT
Crimpergirl might have something substantive to add here.

I read the NYTimes magazine article on recidivism and thought that the argument brought up, that the algorithms used resulted in biased sentencing, to be credible.

Because we do not have a quantitative fundamental model of human behavior there is not way to check if these (or any other) models correctly predict the behavior of the individuals involved. We rely, instead, on a sort of "empirical" model based on supposed observed behavior, this is the basis of "big data" where looking for trends in all the information available is supposed to overcome the need for a "fundamental theory."

The problem is that there is no way to check what is causing the trends. For instance, the trends in the pre-great-recession mortgage boom were explained away, in spite of the fact that it was obviously not sustainable. Yet people who should have known way better discounted their "fundamental knowledge" and went with the trend. The result was this huge economic calamity, the result of using the "empirical data" without understanding it.

In the case of various algorithms meant to "predict" human behavior, for instance the risk a lender takes in offering a loan to some person, or that an insurance company takes in providing insurance as some rate, are built not on some fundamental theory but on historic trends.

If the historic trend is determined by data, and that data is biased, the resulting algorithms will be biased.

Scientists have the same problems, but is is common to publish both the algorithms and the data used in developing the algorithms, and reaching the conclusions. In this way other scientists can check, in detail, how a calculation is done and how well the algorithms predict, independently.

Oddly, the social algorithms in question are not available to the wider public, so it is difficult to assess how well they do. This is very unfortunate for those that are affected by the results of those algorithms, but perhaps in some cases they can resort to a discovery process that has been opened up by recent legislation.

Interestingly, many companies object to the scientific findings of the government. A Congress sympathetic to these complaints has required that the scientific data, and the algorithms, and essentially everything else that is used in scientific finding pertinent to making regulations, be made publicly available within the constraints of law (e.g. personal information has to preserve the confidentiality, say in medical trials patient records can't be associated with the actual person).

For algorithms used by the government, it must be possible for some plaintiff to insist on access to not only the result of the algorithms, but the actual algorithms and the data that were used to produce them. It would seem to be common sense that this information be available, but it seems that the information enjoys some privileged position of not being revealed for review, especially as it pertains to the judicial system. One has only to look at the forensic science laboratory debacle to see that results provided as "science" on the basis of authority only can fall far short of the standard practice of "science." In many of those cases, the results that were provided fully supported the prosecutors, and had no science behind them at all, or at least none that stood up to scientific scrutiny.

In the case of the "private" sector, algorithms used by banks, insurance companies, etc.. should be subject of laws created to insure full disclosure of the development of those practices. The government has an interest in not only insuring "fair play" but also in making sure that these algorithms do not lead the economy down the garden path to more calamity. The private sector will, no doubt, object, but since they have just been so generously bailed out by that same government, it would seem their objection is very weak.

The use of math, that is making the problem "quantitative" and then using that to predict outcomes, is not in question here, but the methodology of how those predictions are made is very relevant. If you put garbage in, you'll get garbage out... unfortunately this has real consequences to people.
Mungeclimber

Trad climber
Nothing creative to say
Sep 8, 2016 - 09:01pm PT
Ed,

I'm going to verbatim that post to my insurance agency, my health care provider, my LEO issuing my tickets, and my co-workers (but not my spouse :)).

Seriously tho, as usual, well said.

thx,
M
Tom

Big Wall climber
San Luis Obispo CA
Sep 8, 2016 - 09:30pm PT




Flip Flop

climber
Earth Planet, Universe
Sep 8, 2016 - 10:24pm PT
Inputs parameters interpretation something something. You could just ask. Fukin Engineers!
jgill

Boulder climber
The high prairie of southern Colorado
Sep 8, 2016 - 10:34pm PT
Can you see in that physics equation a similarity to continuous compounding?
Ed Hartouni

Trad climber
Livermore, CA
Sep 9, 2016 - 12:51am PT
it depends on the the ket, which is usually complex, but can be real... in which case you have the diffusion equation... a negative diffusion constant gets you compounding! (a positive one exponential decay).

mcreel

climber
Barcelona
Sep 9, 2016 - 01:38am PT
The title WMD is unfortunate in that it suggests that math is a problem, and not the way certain people are using it. Science is already morally questionable to a large group of the population, let's hope math is not next on the list. What will they end up teaching in schools if this trend goes on?

Let's Make America Illiterate Again!

By the way, mathematical models can take into account cooperation and altruism. For examples, search for "cooperative game theory".
jonnyrig

climber
Sep 9, 2016 - 07:05am PT
Wait...




Are we saying that math and science are not altruistic after all? I mean, that whole global warming thing for instance.... or guns....
clinker

Trad climber
Santa Cruz, California
Topic Author's Reply - Sep 9, 2016 - 07:21am PT
71 DMT, isn't your number more in the 59 range?

If you put garbage in, you'll get garbage out... unfortunately this has real consequences to people.

Real consequences are finally biting back at Wells Fargo. My wife went to work at Wells about 9 years ago. She is a people person and was outstanding for this area in reviews and surveys of costumer experience. Then everything changed. Sales of new accounts, services and credit cards were pushed on employees.

She started seeing older, decades long loyal customers coming in with fees showing up for various "new accounts". Many of these people had their life savings invested at Wells Fargo. The sales pressure on the tellers was so intense, unethical methods of opening these accounts grew, particularly among the young tellers.

My wife told it like it was to the sales gurus and was told in return that she was not a "team player". Wells decided that the sales were more important than the customers. A major misinterpretation of the numbers.

She quit. If you bank at Wells, find another bank.

5,300 Wells Fargo employees fired over 2 million phony accounts
by Matt Egan @mattmegan5
September 9, 2016: 8:08 AM ET

Wells Fargo fires 5,300 for creating phony accounts
Everyone hates paying bank fees. But imagine paying fees on a ghost account you didn't even sign up for.
That's exactly what happened to Wells Fargo customers nationwide.
On Thursday, federal regulators said Wells Fargo (WFC) employees secretly created millions of unauthorized bank and credit card accounts -- without their customers knowing it -- since 2011.
Powered by SmartAsset.com

MORTGAGE RATES BY MORTGAGE RATES BY
The phony accounts earned the bank unwarranted fees and allowed Wells Fargo employees to boost their sales figures and make more money.
"Wells Fargo employees secretly opened unauthorized accounts to hit sales targets and receive bonuses," Richard Cordray, director of the Consumer Financial Protection Bureau, said in a statement.
Wells Fargo confirmed to CNNMoney that it had fired 5,300 employees over the last few years related to the shady behavior. Employees went so far as to create phony PIN numbers and fake email addresses to enroll customers in online banking services, the CFPB said.
Related: Who owns Wells Fargo? You, me and Warren Buffett
The scope of the scandal is shocking. An analysis conducted by a consulting firm hired by Wells Fargo concluded that bank employees opened over 1.5 million deposit accounts that may not have been authorized.
The way it worked was that employees moved funds from customers' existing accounts into newly-created ones without their knowledge or consent, regulators say. The CFPB described this practice as "widespread." Customers were being charged for insufficient funds or overdraft fees -- because there wasn't enough money in their original accounts.
Additionally, Wells Fargo employees also submitted applications for 565,443 credit card accounts without their customers' knowledge or consent. Roughly 14,000 of those accounts incurred over $400,000 in fees, including annual fees, interest charges and overdraft-protection fees.
The CFPB said Wells Fargo will pay "full restitutions to all victims."
Related: ATM and overdraft fees top $6 billion at the big 3 banks
Wells Fargo is being slapped with the largest penalty since the CFPB was founded in 2011. The bank agreed to pay $185 million in fines, along with $5 million to refund customers.
"We regret and take responsibility for any instances where customers may have received a product that they did not request," Wells Fargo said in a statement.
Wells Fargo has the highest market valuation among any bank in America, worth just north of $250 billion. Berkshire Hathaway (BRKA), the investment firm run legendary investor Warren Buffett, is the company's biggest shareholder.
Of the total fines, $100 million will go toward the CFPB's Civil Penalty Fund, $35 million will go to the Office of the Comptroller of the Currency, and another $50 million will be paid to the City and County of Los Angeles.
"One wonders whether (the CFPB) penalty of $100 million is enough," said David Vladeck, a Georgetown University law professor and former director of the Federal Trade Commission's Bureau of Consumer Protection. "It sounds like a big number, but for a bank the size of Wells Fargo, it isn't really."
Wells Fargo confirmed to CNNMoney that the 5,300 firings took place over several years. The bank listed 265,000 employees as of the end of 2015.
Related: Barclays fined $109 million for trying to hide a deal with rich clients
"At Wells Fargo, when we make mistakes, we are open about it, we take responsibility, and we take action," the bank said in a memo to employees on Thursday.
The CFPB declined to comment on when the investigation began and what sparked it, citing agency policy. "We don't comment on how we uncover these matters," a spokesman said.
As part of the settlement, Wells Fargo needs to make changes to its sales practices and internal oversight.
Customers are fuming. Brian Kennedy, a Maryland retiree, told CNNMoney he detected an unauthorized Wells Fargo account had been created in his name about a year ago. He asked Wells Fargo about it and the bank closed it, he said.
"I didn't sign up for any bloody checking account," Kennedy, who is 57 years old, told CNNMoney. "They lost me as a banking customer and I have warned family and friends."
"Consumers must be able to trust their banks," said Mike Feuer, the Los Angeles City Attorney who joined the settlement.
Feuer's office sued Wells Fargo in May 2015 over allegations of unauthorized accounts. After filing the suit, his office received more than 1,000 calls and emails from customers as well as current and former Wells Fargo employees about the allegations.
Wells Fargo declined to say when it hired a consulting firm to investigate the allegations. However, a person familiar with the matter told CNNMoney the bank launched the review after the L.A. lawsuit was filed.
Even though the Wells Fargo scandal took place nationally, the settlement with L.A. requires the bank to specifically alert all its California customers to review their accounts and shut down ones they don't recognize or want.
"How does a bank that is supposed to have robust internal controls permit the creation of over a half-million dummy accounts?" asked Vladeck. "If I were a Wells Fargo customer, and fortunately I am not, I'd think seriously about finding a new bank."

clinker

Trad climber
Santa Cruz, California
Topic Author's Reply - Sep 9, 2016 - 07:33am PT

Who owns Wells Fargo?


Who owns Wells Fargo anyway? You, me and Warren Buffett
by Matt Egan @mattmegan5
September 8, 2016: 6:23 PM ET


Wells Fargo fires 5,300 for creating phony accounts
Wells Fargo -- which was slapped with a $185 million fine for creating millions of fake accounts, and charging its customers fees for years -- is also America's most valuable bank.
It is one of the nation's most widely-held stocks.
The bank is even backed by America's most beloved investor, Warren Buffett. The legendary billionaire owns $100 million of Wells Fargo (WFC) stock in his personal account, according to FactSet. His firm Berkshire Hathaway (BRKA) is the bank's biggest shareholder, holding nearly 10% of the stock.
Powered by SmartAsset.com

It's ironic, given the shocking revelations, that the key driver for Buffett's investment in Wells Fargo is the bank's reputation for not being one of Wall Street's bad guys.
"Wells Fargo behaves better than the average big bank. But nobody's perfect," Charlie Munger, Berkshire's vice chairman and Buffett's right-hand man, told CNNMoney last year.
Buffett has himself weighed in, saying, according to the Motley Fool, "What Wells Fargo didn't do is what defines their greatness."
Berkshire first bought Wells Fargo stock in 1989, and has since added more.
Related: 5,300 Wells Fargo employees fired over 2 million phony accounts
Today, Wells Fargo is worth $250 billion -- the most valuable bank in the U.S. by market value. To put that in context, Bank of America (BAC) is valued at $160 billion and JPMorgan Chase (JPM) is worth $243 billion.
Many everyday Americans own shares of Wells Fargo in their retirement accounts. Popular mutual fund managers such as Vanguard, BlackRock (BLK) and Fidelity are the bank's top investors, after Berkshire.
And due to Wells Fargo's size, the bank is a staple of most bank ETFs, the popular form of investing in a basket of stocks.

CHASE on the other hand figured out that alerting their costumers immediately to being overdrawn and allowing them to make a deposit by the end of the day to avoid fees was a good thing. How novel.
HighDesertDJ

Trad climber
Sep 9, 2016 - 07:47am PT
Ed- Thank you for such an informed and thorough post. Really appreciate your perspective.
zBrown

Ice climber
Sep 9, 2016 - 08:01pm PT
Kinda makes you wonder where those 5300 Wells Fargo employees are now.

Did they sign non-disclosures?

How much did they get?

Are they exponentially decaying?

clinker

Trad climber
Santa Cruz, California
Topic Author's Reply - Sep 9, 2016 - 09:36pm PT

Kinda makes you wonder where those 5300 Wells Fargo employees are now.



Here are a few.



Workers tell Wells Fargo horror stories
by Matt Egan @mattmegan5
September 9, 2016: 4:24 PM ET

Wells Fargo fires 5,300 for creating phony accounts
Relentless pressure. Wildly unrealistic sales targets. Employees leaning on family members and friends to open unnecessary bank accounts.
That's how more than a dozen former Wells Fargo employees described the bank's culture to CNNMoney.
Wells Fargo (WFC)has been accused by federal regulators of illegal activity on a stunning level. Authorities say employees at the bank secretly created millions of unauthorized bank and credit card accounts between 2011 and July 2015, allowing the bank to make more money in fees and meet internal sales targets.
Powered by SmartAsset.com

SAVINGS RATES BY SAVINGS RATES BY
Wells Fargo agreed to pay penalties of $185 million and fired 5,300 employees over the last few years related to this illegal activity. The news is rocking the industry and rippling across Wells Fargo's millions of customers nationwide.
Former employees tell CNNMoney that they felt incredible demands from managers to meet sales quotas. The same managers turned a blind eye when ethical and even legal lines were crossed.
"I had managers in my face yelling at me," Sabrina Bertrand, who worked as a licensed personal banker for Wells Fargo in Houston in 2013, told CNNMoney. "They wanted you to open up dual checking accounts for people that couldn't even manage their original checking account."
Currently a middle school teacher, Bertrand said she believes the sales targets were set by managers who were higher up: "The sales pressure from management was unbearable."
The pressure cooker environment is also described in a lawsuit filed by Los Angeles against Wells Fargo in May 2015. The lawsuit says that Wells Fargo's district managers discussed daily sales for each branch and employee "four times a day, at 11 am, 1 pm, 3 pm and 5 pm."
It all stems from Wells Fargo's internal goal of selling at least eight financial products per customer. It's what Wells Fargo calls the "Gr-eight initiative." Currently, Wells Fargo boasts an average of about six financial products per customer.
In pursuit of this goal, Wells Fargo employees engaged in all kinds of sordid practices. One of them was internally called "pinning," where the bank issued ATM cards and assigned PIN numbers without customer authorization.
The bankers would impersonate their customers and "input false generic email addresses such as 1234@wellsfargo.com, noname@wellsfargo.com, or none@wellsfargo.com to ensure the transaction is completed," the lawsuit said. The customer, meanwhile, remained completely unaware of the unauthorized activity.
Related: Do more heads need to roll at Wells Fargo?
wells fargo fake account employee
Anthony Try, who worked at Wells Fargo's branches in San Francisco and San Diego as a personal banker and sales representative, told CNNMoney that he believes "management was fully aware of this," but his bosses deliberately "turned a blind eye."
Try, who quit in 2013 because he no longer believed in what he was doing, thinks the illegal activity was systemic.
"It was ingrained in the culture for a long time," he said.
Try, currently a musician and manager, said he did not open unauthorized accounts. However, he did open accounts for friends and family -- with their permission -- in order to meet the incredible demands from managers.
"There would be days where we would open five checking accounts for friends and family just to go home early," he said.
Related: 5,300 Wells Fargo employees fired over 2 million phony accounts
The California lawsuit supports these claims. Wells Fargo paid $50 million to the City and County of Los Angeles to settle the suit as part of the broader $185 million in fines announced on Thursday.
Employees who failed to meet their daily goals were "reprimanded and told to do whatever it takes to meet individual sales quotas," the California lawsuit alleges.
Management even encouraged employees to "achieve 'solutions' through family members," the suit says.
Some Wells Fargo employees say they have "tapped out every family member and friend for accounts," while others say they "spend holiday dinners trying to convince" them to sign up, the lawsuit said.
Employees and the California lawsuit both allege that higher-ups at Wells Fargo also share in the blame for the fraud.
Wells Fargo has "known about and encouraged these practices for years," the California lawsuit said. "Wells Fargo has engineered a virtual fee-generating machine, through which its customers are harmed, its employees take the blame, and Wells Fargo reaps the profits."
"The culprit in this case in not just the individuals involved, but the corporate culture itself," said Julie Ragatz, director of the Center for Ethics in Financial Services at the American College of Financial Services.
In response to CNNMoney's reporting, a Wells Fargo spokeswoman said the "majority of our team members do work hard to do what's in customers' best interest."
Wells Fargo also said it made a number of "fundamental changes to help ensure our team members are not being pressured to sell products."
Those changes include enhanced training that values ethics and how to report concerns, increased risk management professionals at branches and additional mystery shoppers.
Related: Who owns Wells Fargo anyway? You, me and Warren Buffett
One former Wells Fargo employee, who requested his name not be used so he doesn't hurt his career in banking, said he experienced this firsthand. He said managers told him to open unauthorized accounts and, when customers called, to apologize and say it was a mistake.
"This was not done by employees trying to hit their sales numbers, it was more of threats from upper management," he said, adding that workers feared they would lose their job.
According to the California lawsuit, Wells Fargo employees for years engaged in practices known as "gaming." One of them was "sandbagging," which occurred when bankers refused to open accounts requested by customers until the next reporting period to boost their sales quotas, the lawsuit claimed.
A former longtime Wells Fargo consumer bank employee in the Minneapolis region, who also spoke on the condition of anonymity, described a "cutthroat" environment that caused employees to fear for their job and make "bad ethical choices."
"It was a real s**t show over there," he said.

Do the math, heads gotta roll.
rgold

Trad climber
Poughkeepsie, NY
Sep 10, 2016 - 10:04am PT
Ed's discussion is a model for what intelligent discourse ought to look like, which means, in today's superheated post-factual politicized environment, that only an insignificant minority will actually pay attention.

Dr. Rawlins certainly doesn't help by using a preposterous claim as her title. Math is no more or less racist than string beans; the adjective simply can't be applied to such antecedents, and Dr. Rawlins is as well-qualified as anyone to know that. The title only helps to energize a group of radical know-nothings (both the left and the right have growing contingents of such people) who will be happy to parrot the fact that a Harvard-educated mathematician claims that math is racist, without attending to any of the far more intelligent details of Dr. Rawlins' arguments. Sure as shootin' some benighted school board somewhere will try to reduce the math requirements for students with Dr. Rawlins as a reference.

A flaw in Dr. Rawlins' approach is that the proposition that data science is racist cannot be confirmed or disconfirmed by using results of data science. But without appropriate statistics, she has no argument. If math is truly racist (which she doesn't actually believe for even a second), then it would be the height of both hypocrisy and logical invalidity to employ a discredited tool to advance her own position.

Don't get me wrong, there can be major problems in trying to use the past to predict the future, and the interest of medical, commercial, and governmental organizations in statistical outcomes takes probabilistic reasoning for groups and applies it to individuals in ways that can easily be unjust.

So I'm not questioning some of Dr. Rawlins' conclusions, just her use of absurd attention-getting headlines. Joining the sensationalists might get your blog more eyeballs, but in my opinion your integrity takes a hit for that.
rbord

Boulder climber
atlanta
Sep 10, 2016 - 01:41pm PT
Thanks Ed. Nicely said.

If you train a system using racist data, you're going to get a racist system. Like the one we've been trained for.

I think the thinking mistake we make is to believe that what we do in our heads - our thinking - is objective. We believe that what we're doing in our heads is just math, when I think really what we're doing is survivor-biased math. Our belief creation processes evolved (i.e. have been trained) to work to our advantage, not (necessarily) to work to the advantage of objective mathematical thinking.
Bad Climber

Trad climber
The Lawless Border Regions
Sep 10, 2016 - 02:29pm PT
Good pts., Rgold. I was aching for some stats of any kind as support. Like you, I'm not saying she's wrong, but evidence is nice. For example, with the algorithm for recidivism, was there any positive outcome? How does sentencing influence recidivism anyway? Can we see a before/after comparison with the algorithm?

BAd
Ed Hartouni

Trad climber
Livermore, CA
Sep 10, 2016 - 02:47pm PT
http://fivethirtyeight.com/features/prison-reform-risk-assessment/
JEleazarian

Trad climber
Fresno CA
Sep 19, 2016 - 12:50pm PT
So I'm not questioning some of Dr. Rawlins' conclusions, just her use of absurd attention-getting headlines. Joining the sensationalists might get your blog more eyeballs, but in my opinion your integrity takes a hit for that.

Exactly. There's an additional problem, to which Ed refers, and in which I've spent a considerable portion of my adult life. Classical statistical theory assumes that we properly identified the components of the model. We may then use the normal statitical tools (e.g. run regressions) to determine the values of the parameters and their statistics of fit.

Statistical models used for the soical sciences, however, and where I live in economics in particular, usually involve attempting to make statistical inferences with non-experimental data. Among other things, we attempt to prove not only the value of model parameters, but whether a putative explanatory variable even belongs in the model. Classical statistical theory will overestimate the importance and the statistical significance of some variables in that situation, and does a much poorer jub than Bayesian theory in that regard.

The Bayesians recognized that people have prior opinions, and that additional data will cause people to modify those opinions, but not necessarily to jettison them. Of course, since Bayesian results depend on the prior distribution employed, they don't provide as certain a result. I find that increased uncertainty, however, a much more honest evaluation of what we know.

So back to the original post. Math (really statistical analysis) has no more racism than gravity. Its users, however, have biases that their results reflect.

John
Tom Turrentine

Trad climber
Santa Cruz
Sep 19, 2016 - 02:33pm PT
I'm not a statistician, but work around a lot of statistical modelers in transport research. Seems the models work pretty well as long as nothing is changing, but change technology for example, and their ability to predicts drop. Radical changes in automobile-networks etc.. technology is making prediction less reliable.

I like the book by Nate Silver, the Signal and the Noise. What I learned from that was as long as he had 100 years of data, like baseball, he could make some pretty good predictions for example about left handed pitchers who throw sidearm, but if he only had a few years of data, no way.

jgill

Boulder climber
The high prairie of southern Colorado
Sep 19, 2016 - 03:15pm PT
Nice post, John E. My father was a Bayesian in the areas of business and economics.
Messages 21 - 40 of total 67 in this topic << First  |  < Previous  |  Show All  |  Next >  |  Last >>
Return to Forum List
 
Our Guidebooks
spacerCheck 'em out!
SuperTopo Guidebooks

guidebook icon
Try a free sample topo!

 
SuperTopo on the Web

Recent Route Beta