Beyond Fairness: Reparative Algorithms to Address Historical Injustices of Housing Discrimination in the US (original) (raw)
Related papers
Berkeley Business Law Journal, Vol. 21(1), 2024
Credit discrimination undermines consumer financial autonomy and distorts market pricing of lending risks. To ensure equal access to credit, existing federal fair lending laws—e.g., Equal Credit Opportunity Act, Fair Housing Act—prohibit lenders from considering race, sex, age, or national origin in their lending decisions. For decades, the fair lending laws have largely held the banking industry in check. However, as lenders increasingly delegate lending decisions to artificial intelligence (AI) through the service of fintech and data intermediaries, it is questionable whether existing laws can still adequately safeguard equal credit access. This Article argues that the current fair lending regime can no longer protect consumers in the age of AI. This is because our regime does not account for harms traceable to automatic, unsupervised algorithmic processes. Unlike human actors, algorithms cannot desire to cause harm or intend to use suspect factors. Yet, courts and litigants are constrained by the language of the fair lending laws to hold AI accountable under an antiquated legal theory—treating discrimination as analogous to common law torts. Under this regime, victims of AI discrimination carry the burden of showing lender animus and causal explanations linking the victim’s injury to the lender’s specific acts or policies. Consequently, such victims are often barred from recovery due to insurmountable pleading and evidentiary hurdles. Thus, any attempt to combat AI discrimination must consider two unique features of algorithmic harm. First, an algorithm’s discriminatory decision may have no explicable connection—let alone causal relation—to the acts or policies of the lender due to the algorithm’s self-learning capabilities. Second, whether an algorithm discriminates depends on a host of variables typically outside the lenders’ control. The unpredictable nature of AI calls into question the effectiveness of regulating AI bias under the fair lending laws—a conduct-based liability regime that emphasizes causation, reasonable foreseeability, and ex-ante risk mitigation. As a blueprint for reform, this Article proposes an alternative harm-based framework to address the root cause of AI discrimination: data opaqueness. To implement this framework, this Article recommends the CFPB to adopt a new rule prohibiting the use of “black box” algorithms in consumer lending, pursuant to the CFPB’s authority to prohibit “unfair, deceptive, or abusive acts and practices” (UDAAPs) under the Dodd-Frank Act.
SSRN Electronic Journal
In respect to racial discrimination in lending, we introduce global Shapley value and Shapley-Lorenz explainable AI methods to attain algorithmic justice. Using 157,269 loan applications during 2017 in New York, we confirm that these methods, consistent with the parameters of a logistic regression model, reveal prima facie evidence of racial discrimination. We show, critically, that these explainable AI methods can enable a financial institution to select an opaque creditworthiness model which blends out-of-sample performance with ethical considerations.
Algorithmic discrimination in the credit domain: what do we know about it?
AI & SOCIETY
The widespread usage of machine learning systems and econometric methods in the credit domain has transformed the decision-making process for evaluating loan applications. Automated analysis of credit applications diminishes the subjectivity of the decision-making process. On the other hand, since machine learning is based on past decisions recorded in the financial institutions’ datasets, the process very often consolidates existing bias and prejudice against groups defined by race, sex, sexual orientation, and other attributes. Therefore, the interest in identifying, preventing, and mitigating algorithmic discrimination has grown exponentially in many areas, such as Computer Science, Economics, Law, and Social Science. We conducted a comprehensive systematic literature review to understand (1) the research settings, including the discrimination theory foundation, the legal framework, and the applicable fairness metric; (2) the addressed issues and solutions; and (3) the open chall...
William Mary Business Law Review, 2015
For decades the agencies charged with minding the 'fair credit and lending' shop turned a blind eye to those (lenders) who pilfered minority homeownership (and consequently minority wealth) by extending mortgage lending products that were, in many cases, unequal to similarly situated non-minority counterparts. Since the 1950s, when the federal government endorsed homeownership policies for minorities, and the 1960s, when antidiscriminatory lending laws were enacted, access to fair mortgage credit has been unattainable. Unbridled lending discrimination culminated in massive foreclosures for a disproportionate number of minority homeowners during the Housing and Foreclosure Crisis. Lenders disparately foreclosed upon upper class, middle class and lower class minority homeowners. The effect of these foreclosures widened homeownership gaps between whites and minorities. Foreclosures were more prevalent for minority homeowners regardless of economic class. Lending discrimination, and subsequent forfeiture of homes, undoubtedly altered the perception of the American Dream, and resulted in losses of generational wealth for minorities, furthered racial segregation and prolonged the stagnancy of the real estate market. Unquestionably then, lending discrimination is not a minority problem, but is an American problem. Therefore, agencies with jurisdiction to enforce lending and credit laws must, first, duly enforce these laws and, second, create civil or criminal mechanisms that effectively and finally eliminate unfair lending.
Riding the Stagecoach to Hell: A Qualitative Analysis of Racial Discrimination in Mortgage Lending
Recent studies have used statistical methods to show that minorities were more likely than equally qualified whites to receive high-cost, high-risk loans during the U.S. housing boom, evidence taken to suggest widespread discrimination in the mortgage lending industry. The evidence, however, was indirect, being inferred from racial differentials that persisted after controlling for other factors known to affect the terms of lending. Here we assemble a qualitative database to generate direct evidence of discrimination. Using a sample of 220 statements randomly selected from documents assembled in the course of recent fair lending lawsuits, we code texts for evidence of individual discrimination, structural discrimination, and potential discrimination in mortgage lending practices. We find that 76 percent of the texts indicated the existence of structural discrimination, with only 11 percent suggesting individual discrimination alone. We then present a sample of texts that were coded as discriminatory to reveal the way in which racial discrimination was embedded within the social structure of U.S. mortgage lending, and to reveal the specific microsocial mechanisms by which this discrimination was effected.
Black Loans Matter: Distributionally Robust Fairness for Fighting Subgroup Discrimination
ArXiv, 2020
Algorithmic fairness in lending today relies on group fairness metrics for monitoring statistical parity across protected groups. This approach is vulnerable to subgroup discrimination by proxy, carrying significant risks of legal and reputational damage for lenders and blatantly unfair outcomes for borrowers. Practical challenges arise from the many possible combinations and subsets of protected groups. We motivate this problem against the backdrop of historical and residual racism in the United States polluting all available training data and raising public sensitivity to algorithimic bias. We review the current regulatory compliance protocols for fairness in lending and discuss their limitations relative to the contributions state-of-the-art fairness methods may afford. We propose a solution for addressing subgroup discrimination, while adhering to existing group fairness requirements, from recent developments in individual fairness methods and corresponding fair metric learning ...
Lessons Learned: Statistical Techniques and Fair Lending
There remains strong concern that, even after several years of intense scrutiny, lending discrimination persists. The concerns encompass issues of lending denial disparities, use of predatory lending tactics, and potential disparate impact arising from increased use of credit scoring. This article specifically addresses disparate treatment of loan applications by analyzing data collected in fair lending examinations conducted at national banks during the period 1994 through mid-1999. This information will be useful to banks interested in monitoring their performance, to consumers interested in determining factors that influence their ability to purchase homes, and to policy makers concerned with discrimination issues.
The Social Structure of Mortgage Discrimination
Housing studies, 2018
In the decade leading up to the U.S. housing crisis, black and Latino borrowers disproportionately received high-cost, high-risk mortgages-a lending disparity well documented by prior quantitative studies. We analyze qualitative data from actors in the lending industry to identify the social structure though which this mortgage discrimination took place. Our data consist of 220 depositions, declarations, and related exhibits submitted by borrowers, loan originators, investment banks, and others in fair lending cases. Our analyses reveal specific mechanisms through which loan originators identified and gained the trust of black and Latino borrowers in order to place them into higher-cost, higher-risk loans than similarly situated white borrowers. Loan originators sought out lists of individuals already borrowing money to buy consumer goods in predominantly black and Latino neighborhoods to find potential borrowers, and exploited intermediaries within local social networks, such as co...
Discrimination in the Credit and Housing Markets: Findings and Challenges
Handbook on the Economics of Discrimination, 2006
Economics all too seldom provides straightforward guidelines for designing and analyzing statistical materials on subjects of great social importance. Since the economic theory of discrimination does provide a simple approach, it is too bad that studies of whether banks discriminate in mortgage lending have not utilized these insights.-Gary Becker (1993, P. 18) Eventually, even the definition of discrimination comes to mean different things to blacks and whites.-Derrick Bell (1980, P. 658)