The hidden rules of the credit score game

Algorithmic bias

Experts of one of the world leading big law firms, White & Case, explained in their paper Algorithms and bias: What lenders need to know that even clearly unintentional algorithms direct financial technologies may lead to discriminative decisions. Why? Creditors and lenders have access now, in the age of Big Data, to so-called nontraditional data, such as Internet activity, shopping patterns and other data which are not necessarily directly related to creditworthiness. These data are analyzed by using the techniques of now very popular machine learning.

Traditional algorithms use the rules of arithmetics and logic defined by the designer of the algorithm. Say, IF Borrower payed back her previous credits without any delay THEN increase her credit score by X points. But  machine learning  techniques relies don’t have a previously defined algorithm, they generate algorithms based on patterns found in large datasets. Take for example the approval of loan request. The software has stored analyzed data of financial behavior of many thousands of previous customers. Loading the credit history of a new applicant as input data, a machine learning algorithm might calculate the output, something like the probability that the applicant will default.

There are justified concerns that algorithms might bring biased, and may be unfavorable decisions for minorities. Ideally, the decision makers should take into account the data permitted by ECA for the borrower only. But we live in networks surrounded by our neighbors, friends, and peers etc. If the creditors analyze data of your social network friends to rate your creditworthiness, it might lead to discrimination based on data creditors are not permitted to consider. Living address might matters a lot, and ZIP code is considered a dangerous variable, and the term redlining expresses the discriminatory practice (historically for people lived in black inner city neighborhoods).

A machine learning algorithms may find that there is a correlation between your creditworthiness and the financial behaviour of your friends or neighbors. It is a complicated situation: a creditor cannot deny your request on the basis that many of your friends delayed to repay their loan. Also, they should explain the basis of their credit request denial. But if nontraditional data are used, it is very difficult to give transparent and understandable explanation.

It is understandable that data from your social network cannot be used for evaluating your financial future. However, we choose many our future activities based on recommendation systems. These recommendations influence our choices of hotels, restaurants, dating partners, movies, just to give an unranked list.
The recommender algorithms use data also about ”stuffs what my friends like“. We will discuss in more detail in subsection Sec:10.1.

Should we like algorithms? If you are ready to answer the question with ”No”, did you think whether or not we better off by returning to the personality-based subjective credit evaluation?

In May, 2016, the Obama administration’s The Treasury Department issued a white paper titled,  Opportunities and Challenges in Online Marketplace Lending. In addition to the traditional players, online marketplace lending companies has been formed emerged offering faster credit for consumers and small businesses. It was a good news that the Treasury Department found important to analyze the opportunities and risks presented of this new type of crediting system.

Towards fair algorithms?

Computer scientist realized that algorithms might lead even unintentionally to discrimination. Why? Data mining methods are based upon assumptions that comes from the pioneers of modern science, as Galileo, Kepler and Newton: looking into the data from the past implies our ability to predict the future. While this method worked wonderfully to predict the motion of the planets, should we also assume that historical data for the social behavior are useful for prediction?

There are now algorithms for forecasting crimes based on historical data. Patterns for time of the day, seasonality, weather, location (vicinity of bars, bus bus stops factors), crime level in the past and similar data help police departments to have a better chance to prevent potential crimes. As always, while the goal of \emph{predictive policing} promises to be race-neutral and objective, there are also justified concerns that the application of algorithmic approach leads to the emergence of new problems related security, privacy, and constitutional rights of citizens. (A.G. Ferguson: \emph{The Rise of Big Data Policing Surveillance, Race, and the Future of Law Enforcement}. Again: algorithms behind predictive policing – much more often than not – help the work of law enforcement, but they are not silver bullets to eliminate crimes. As a Lithuanian data scientist Indrė Žliobaitė, now in the University of Helsinki in Finland writes in a position paper about ”Fairness-aware machine learning”: ”usually predictive models are optimized for performing well in the majority of the cases, not taking into account who is affected the worst by the remaining inaccuracies”. It is a very difficult question, we all know that there are human faces and fates behind the numbers.

We know the horror stories, and I decided not to repeat them, when intentionally neutral algorithms produced sexist or racist output. One reason is that machines learn by examples extracted from the data. Data generated by humans reflect human bias. Yes, it may happen that algorithms may sustain prejudice or maintain social hierarchy.

Social scientists and computer scientist should cooperate to generate ”ethical algorithms”. Ethics, i.e. moral philosophy investigates what is ”good” or ”bad” behavior. (I leave the answers for the philosophers.) From the perspective of machine learning the question is how to train algorithms to teach them to bring moral decisions. Than date can be preprocessed, and unethical data could be eliminated. We may expect that lot of studies will be conducted to understand the scope and limits of building ethical algorithms, and we should accept that ”fairness” is far to be a well-defined concept.

Beyond the algorithms: the Lending Circles and the credit score game

Not only algorithms, people also can learn :-). As a researcher now at Univetsity of Arizona, Mark Kear describes and analyzes an example how immigrants learned that (i) they should play the credit score game, (ii) it is possible to improve their credit history. Kear was a participant and observer of a Lending Circle. Lending Circles are organized by the Mission Asset Fund (MAF), a San Francisco-based nonprofit organization to help increase the credit scores of low-income families. People learn strategies to report data, which improves their creditworthiness. MAF managed to increase the credit scores significantly (with 168 in one case study).

Instead of summary

As John von Neumann wrote in his paper Can we survive technology?:

All experience shows that even smaller technological changes than those now in the cards profoundly transform political and social relationships. Experience also shows that these transformations are not a priori predictable and that most contemporary ‘first guesses’ concerning them are wrong. For all these reasons, one should take neither present difficulties nor presently proposed reforms too seriously. \dots

The one solid fact is that the difficulties are due to an evolution that, while useful and constructive, is also dangerous. Can we produce the required adjustments with the necessary speed? The most hopeful answer is that the human species has been subjected to similar tests before and seems to have a congenital ability to come through, after varying amounts of trouble. To ask in advance for a complete recipe would be unreasonable. We can specify only the human qualities required: patience, flexibility, intelligence.







How a credit score is calculated, and how objective it is?

The main goal of this book to uncover the hidden rules behind our navigation between subjectivity and objectivity. We cannot deny (and I don’t have any intention to do), that algorithms are based on human’s assumptions. After the assumptions are made, the evaluation is the outcome of an automatic procedure). To set a credit score algorithms the first question to decide is the input data to be taken into account. FICO uses five factors: (i) the history how did you pay your bills; (ii) how much money you owe on credit cars. mortgages, loans etc.; (iii) the length of your credit history (the longer the better); (iv) mix of credit (the more divers the better); (v) new credit applications (don’t open too many new accounts too fast). The next natural questions is whether of not is reasonable assume that all the five factors have the same importance? Assuming the answer is yes, to each input variable we should assign 20\% weight. It is more plausible to assume that there are more and less important factors, and FICO uses the following weights:

  • Payment history: 35%
  • Amounts owed: 30%
  • Length of credit history: 15%
  • Credit mix in use: 10%
  • New credit: 10%

We already know what are the factors the calculations take into account, and it is very crucial to know  factors which don’t count.  The Equal Credit Opportunity Act (ECA), doesn’t allow creditors in the United States to discriminate based on race, color, religion, national origin, sex, marital status, age.

There are variations how to calculate the credit score. Somewhat more technically speaking, the credit score (i.e a single number is the output of the algorithm), and the simplest way to get is by summing the weighted inputs.  FICO uses a scale which runs from 300 to 850. As a blogger writes: “FICO should disclose what goes into its all important algorithms. They say they don’t want people to game them, but considering their importance in buying a house or a car, it can’t be a black box, that only FICO knows”. Figuere shows the national distribution of the FICO scores.

Stay tuned! next post: Struggle for fairness


Credit scores: a little history

Buy, buy, buy! We want to buy even we cannot afford it. Therefore if you need money you should ask somebody to lend you. This ”somebody” (whoever she is, your friend
or a bank) has a single question: ”May I trust the borrower will repay the loan?” In a world when people did not live in the ocean of data the potential lenders characterized qualitatively the borrowers. “He looks a nice, reliable guy, well, I think he will repay… in addition, he promised to pay this and this percent of interest”. Owners of corner grocery stores developed skills for century to classify clients, as reliable and not reliable.  I find interesting, but not surprising that the oldest credit reporting agency in the USA emerged from the grocery business. Cator Woolford was a a grocer in Chattanooga, Tennessee. He collected data from his customers, produced a book, and sold copies of the book to the local Retail Grocer’s Association. Based on his success, together with his younger brother Guy, a lawyer, opened in Atlanta a very small business, they called ”Retail Credit Company”. This small business evolved what we call now Equifax Inc. (one of the three giant consumer credit bureaus, the other two are Experian, and TransUnion), which
collects and process information over 800 million individual consumers.

When people should judge other people’s character, it is truly truly subjective. Granting or denying loans or credit requests was very from being objective, and age-, gender- or race-based discriminations happened again and again. To help the decision makers by quantitative analysis was a big step towards objectivity. William R Fair (1923-1996) and Earl Isaac (1921-1983) started to build mathematical models for predicting the behaviour of the potential borrowers. An initial version of a Credit Application Scoring Algorithm was introduced in 1958. This algorithm generated three possible behaviors: the borrower will pay on time, will pay with delay, or not pay at all. The Fair Isaac Corporation has been established, and developed and algorithm and software to calculate what became the famous/infamous FICO score.

Stay tuned! (How a credit score is calculated, and how objective it is?}

Lists in a brain game

Our brain processes external information perceived by all of our sensory systems. The incoming information is useful only if we are able to comprehend, and lists help to organize new information. There are situations, when people are in a complex, dynamic
environment and they should rapidly understand ”what is happening”, and they should bring decisions followed by actions. Historically, military command and control is a field from where the theory and practice of situated awareness emerged. However, other activities, such as air traffic control, fire fighting, aviation and more ordinary complex tasks, as driving a car or even riding a bicycle needs to comprehend the rapidly the changing environment and react. Situation awareness starts with the perception of environmental elements and events with respect to time or space, followed by the comprehension of their meaning, and by the  projection of possible future events.

Lists help to comprehend incoming information. Kim’s game is a famous example how a
complex environments should be mapped into a list, and how to improve the efficiency of the the comprehension. Bert and Kate McKay, founders of the Art of Manliness summarized the origin of the game so nicely, that I should copy here:

”In Rudyard Kipling famous novel Kim, Kimball O’Hara, an Irish teenager, undergoes
training to be a spy for the British Secret Service. As part of this training, he is
mentored by Lurgan Sahib, an ostensible owner of a jewelry store in British India,
who is really doing espionage work against the Russians. Lurgan invites both his boy servant and Kim to play the “Jewel Game.” The shopkeeper lays 15 jewels out on a tray, has the two young men look at them for a minute, and then covers the stones with a newspaper. The servant, who has practiced the game many times before, is easily able to name and exactly describe all the jewels under the paper, and can even accurately guess the weight of each stone. Kim, however, struggles with his recall and cannot transcribe a complete list of what lies under the paper. Kim protests that the servant is more familiar with jewels than he is, and asks for a rematch. This time the tray is lined with odds and ends from the shop and kitchen. But the servant’s memory easily beats Kim’s once again, and he even wins a match in which he only feels the objects while blindfolded before they are covered up. Both humbled and intrigued, Kim wishes to know how the boy has become such a master of the game. Lurgan answers: “By doing it many times over till it is done perfectly — for it is worth doing.”

Umberto Eco (1932-2016), the celebrated Italian novelist and and a public intellectual famously wrote: ”We like lists because we don’t want to die”, and list are means of grasping the incomprehensible.

Stay tuned! more to come!


Ranking Oprah

simply from Wikipedia


Winfrey at the White House for the 2010 Kennedy Center Honors

Winfrey was called “arguably the world’s most powerful woman” by CNN and,[105] “arguably the most influential woman in the world” by The American Spectator,[106] “one of the 100 people who most influenced the 20th Century” and “one of the most influential people” from 2004 to 2011 by TIME. Winfrey is the only person in the world to have appeared in the latter list on ten occasions.[107]

At the end of the 20th century, Life listed Winfrey as both the most influential woman and the most influential black person of her generation, and in a cover story profile the magazine called her “America’s most powerful woman”.[108] In 2007, USA Todayranked Winfrey as the most influential woman and most influential black person of the previous quarter-century.[109] Ladies Home Journal also ranked Winfrey number one in their list of the most powerful women in America and Senator Barack Obama has said she “may be the most influential woman in the country”.[110] In 1998, Winfrey became the first woman and first African American to top Entertainment Weekly‘s list of the 101 most powerful people in the entertainment industry.[111] Forbes named her the world’s most powerful celebrity in 2005,[112] 2007,[113] 2008,[101] 2010,[114] and 2013.[115] As chairman of Harpo Inc., she was named the most powerful woman in entertainment by The Hollywood Reporter in 2008.[116] She has also been listed as one of the most powerful 100 women in the world by Forbes, ranking fourteenth in 2014.[117] In 2010, Life magazine named Winfrey one of the 100 people who changed the world, alongside such luminaries as Jesus ChristElvis Presley and Lady Mary Wortley Montagu. Winfrey was the only living woman to make the list.[118]

Columnist Maureen Dowd seems to agree with such assessments: “She is the top alpha female in this country. She has more credibility than the president. Other successful women, such as Hillary Clinton and Martha Stewart, had to be publicly slapped down before they could move forward. Even Condi has had to play the protegé with Bush. None of this happened to Oprah – she is a straight ahead success story.[119] Vanity Fair wrote: “Oprah Winfrey arguably has more influence on the culture than any university president, politician, or religious leader, except perhaps the Pope.[120] Bill O’Reilly said: “this is a woman that came from nothing to rise up to be the most powerful woman, I think, in the world. I think Oprah Winfrey is the most powerful woman in the world, not just in America. That’s – anybody who goes on her program immediately benefits through the roof. I mean, she has a loyal following; she has credibility; she has talent; and she’s done it on her own to become fabulously wealthy and fabulously powerful.”[121]

In 2005, Winfrey was named the greatest woman in American history as part of a public poll as part of The Greatest American. She was ranked No. 9 overall on the list of greatest Americans. However, polls estimating Winfrey’s personal popularity have been inconsistent. A November 2003 Gallup poll estimated that 73% of American adults had a favorable view of Winfrey. Another Gallup poll in January 2007 estimated the figure at 74%, although it dropped to 66% when Gallup conducted the same poll in October 2007. A December 2007 Fox News poll put the figure at 55%.[122] According to Gallup’s annual most admired poll, Americans consistently rank Winfrey as one of the most admired women in the world. Her highest rating came in 2007[123] when she was statistically tied with Hillary Clintonfor first place.[124] In a list compiled by the British magazine New Statesman in September 2010, she was voted 38th in the list of “The World’s 50 Most Influential Figures 2010”.[125]

In 1989, she was accepted into the NAACP Image Award Hall of Fame

Reputation and Ranking systems

Gloria Origgi is an Italian philosopher and cognitive scientist working now at the famous EHESS (CNRS) in Paris. Her book Reputation: What It Is and Why It Matters has been published now by the Princeton University Press .

A compelling exploration of how reputation affects every aspect of contemporary life

Reputation touches almost everything, guiding our behavior and choices in countless ways. But it is also shrouded in mystery. Why is it so powerful when the criteria by which people and things are defined as good or bad often appear to be arbitrary? Why do we care so much about how others see us that we may even do irrational and harmful things to try to influence their opinion? In this engaging book, Gloria Origgi draws on philosophy, social psychology, sociology, economics, literature, and history to offer an illuminating account of an important yet oddly neglected subject.

Origgi examines the influence of the Internet and social media, as well as the countless ranking systems that characterize modern society and contribute to the creation of formal and informal reputations in our social relations, in business, in politics, in academia, and even in wine. She highlights the importance of reputation to the effective functioning of the economy and e-commerce. Origgi also discusses the existential significance of our obsession with reputation, concluding that an awareness of the relationship between our reputation and our actions empowers us to better understand who we are and why we do what we do.

Compellingly written and filled with surprising insights, Reputation pins down an elusive subject that affects everyone.”

I hope to have the book in my mail box when I will be back from Budapest to Kalamazoo,  Michigan.

Rating graduate students applicants

December. As a college professor, my seasonal duty is writing letters of recommendations, and rate students based on  several criteria to help them to be accepted by graduate schools. Students should ask a number of professors to evaluate them. Occasionally I have to tell a student that I would not be able to be a strong recommender, so it is better not to ask me. We, evaluators, combine quasi-objective data (say, grades) and subjective impressions to generate a rating score. Subjectivity is far from being identical to random, and college professors don’t have better ways to help students and graduate programs to find a good match.  Admission committee has a strong interest in ensuring they only accept mature, polite, reliable and stable people into their program, and my professional duty is to help them.

CollegeNET  is a corporation, which provides software as a service for many universities, among others for admissions and application evaluation. There are six criteria to rate students:

  • Knowledge in chosen field
  • Motivation and perseverance toward goals
  • Ability to work independently
  • Ability to express thoughts in speech and writing
  • Ability/potential for college teaching
  • Ability to plan and conduct research

We should choose among five options: Exceptional (Upper 5%) Outstanding (Next 15%) Very Good (Next 15%) Good (Next 15%)  (Next 50.)

(In some other softwares the “exceptional” is the upper 2%. I noticed that while I am ready to place students in exceptional category if is defined as upper 5%, and very infrequently, if they should be in the upper 2%.)

How  do we generate the numbers and choose the appropriate rubric? In principle, a micro-rationalist, bottom up approach would work: teachers could collect and store data from students back to decades, and they might have a formal algorithm to calculate the percentages. I do believe, still many of us adopts top down strategies. I ask myself: do I want to grant an “all exceptional” set of grades? Does the applicant  have a clearly weakest point, so  should I check the third or maybe the fourth rubric?  How about to check four exceptional and two outstanding rubrics?
Good or bad, decision makers calculate the sum of the grades, analyze the grade distribution.

As Churchill could have told: Quantification is the worst form of evaluation, except for all the others.