The shortcomings of causal and counterfactual thinking about racial discrimination

Legal claims of disparate impact discrimination go something like this: A company uses some system (e.g., hiring test, performance review, risk assessment tool) in a way that impacts people. Somebody sues, arguing that it has a disproportionate adverse effect on racial minorities, showing initial evidence of disparate impact. The company, in turn, defends itself by arguing that the disparate impact is justified: their system sorts people by characteristics that—though incidentally correlated with race—are relevant to its legitimate business purposes. Now, the person who brought the discrimination claim is tasked with coming up with an alternative—that is, a system with less disparate impact and still fulfills the company’s legitimate business interest. If the plaintiff finds such an alternative, it must be adopted. If they don’t, the courts have to, in theory, decide how to tradeoff between disparate impact and legitimate business purpose.

Much of the research in algorithmic fairness, a discipline concerned with the various discriminatory, unfair, and unjust impacts of algorithmic systems, has taken cues from this legal approach—hence, the deluge of parity-based “fairness” metrics mirroring disparate impact that have received encyclopedic treatment by computer scientists, statisticians, and the like in the past few years. Armed with intuitions closely linked with disparate impact litigation, scholars further formalized the tradeoffs between something like justice and something like business purpose—concepts that crystallized in the literature under the banners of “fairness” and “efficiency.”

An interview with Seda Gürses and Bekah Overdorf

Software that structures increasingly detailed aspects of contemporary life is built for optimization. These programs require a mapping of the world in a way that is computationally legible, and translating the messy world into one that makes sense to a computer is imperfect. Even in the most ideal conditions, optimization systems—constrained, more often than not, by the imperatives of profit-generating corporations—are designed to ruthlessly maximize one metric at the expense of others. When these systems are optimizing over large populations of people, some people lose out in the calculation.

Official channels for redress offer little help: alleviating out-group concerns is by necessity counter to the interests of the optimization system and its target customers. Like someone who lives in a flight path but has never bought a plane ticket complaining about the noise to an airline company, the collateral damage of optimization has little leverage over the system provider unless the law can be wielded against it. Beyond the time-intensive and uncertain path of traditional advocacy, what recourse is available for those who find themselves in the path of optimization?

In their 2018 paper POTs: Protective Optimization Technologies (updated version soon forthcoming at this same link), authors Rebekah Overdorf, Bogdan Kulynych, Ero Balsa, Carmela Troncoso, and Seda Gürses offer some answers. Eschewing the dominant frameworks used to analyze and critique digital optimization systems, the authors offer an analysis that illuminates fundamental problems with both optimization systems and the proliferating literature that attempts to solve them.

POTs—the analytical framework and the technology—suggest that the inevitable assumptions, flaws, and rote nature of optimization systems can be exploited to produce “solutions that enable optimization subjects to defend from unwanted consequences.” Despite their overbearing nature, optimization systems typically require some degree of user input; POTs uses this as a wedge for individuals and groups marginalized by the optimization system to influence its operation. In so doing, POTs find a way to restore what optimization seeks to hide, revealing that what gets laundered as technical problems are actually political ones.

Below, we speak with Seda Gürses and Bekah Overdorf, two members of the POTs team, who discuss the definition of optimization system, the departures the POTs approach makes from the digital ethics literature, and the design and implementation of POTs in the wild.

Three competing theories of money

In the past year, Modern Monetary Theory (MMT) has shifted the policy debate in a way that few heterodox schools of economic thought have in recent memory. MMT’s central notion—that nations with their own strong currencies face no inherent financial constraints—has made its way into politics and, notably, the world of finance. The last few months have brought MMT explainers from financial media outlets including Reuters, CNBC, Bloomberg, Barron’s, and Business Insider, as well as from investment analysts at Wall Street firms including Goldman Sachs, Bank of America, Fitch, Standard Chartered and Citigroup.

Popularizing the shorthand notion that “deficits don’t matter” has been an achievement for those promulgating MMT. Yet one largely unappreciated change brought about by the MMT debates involves a somewhat subtler point: a shift in the implicit story we tell about money.

The rise of MMT poses a challenge to the mainstream commodity money story. That parable, familiar to anyone who has taken high school economics or read Adam Smith, involves an inefficient barter system that gives way to the more convenient use of some token that represents value, typically a precious metal. If government plays a role in this story, it is only to regulate money after the marketplace births it.

The MMT parable—known in the literature as chartalism—reverses the commodity money view. For chartalists, money arises through an act of law, namely the levying of a tax which requires citizens to go out and get that which pays taxes; the state comes first and markets are subsequent. As Abba Lerner puts it, money is “a creature of the state.”

Statistical prediction is increasingly pervasive in our lives. Can it be fair?

The Allegheny Family Screening Tool is a computer program that predicts whether a child will later have to be placed into foster care. It's been used in Allegheny County, Pennsylvania, since August 2016. When a child is referred to the county as at risk of abuse or neglect, the program analyzes administrative records and then outputs a score from 1 to 20, where a higher score represents a higher risk that the child will later have to be placed into foster care. Child welfare workers use the score to help them decide whether to investigate a case further.

Travel search engines like Kayak or Google Flights predict whether a flight will go up or down in price. Farecast, which launched in 2004 and was acquired by Microsoft a few years later, was the first to offer such a service. When you look up a flight, these search engines analyze price records and then predict whether the flight's price will go up or down over some time interval, perhaps along with a measure of confidence in the prediction. People use the predictions to help them decide when to buy a ticket.

An Interview with Mark Granovetter

Few living scholars have had the influence of Mark Granovetter. In a career spanning almost 50 years, his seminal contributions to his own field of sociology have spread to shape research in economics, computer science, and even epidemiology.

Granovetter is most widely known for his early contributions to social network analysis—in particular his 1973 article, “The Strength of Weak Ties.” In that paper, Granovetter demonstrated that, because of the way social networks evolve, “weak ties” between people often form bridges between clusters of more strongly connected individuals and thus serve as important conduits of novel information. This surprising finding has proven to have important and enduring implications for a diverse range of fields. The paper remains one of the most cited social science articles of all time.

Can you fix political problems with new web infrastructures?

The internet's early proliferation was steeped in cyber-utopian ideals. The circumvention of censorship and gatekeeping, digital public squares, direct democracy, revitalized civic engagement, the “global village”—these were all anticipated characteristics of the internet age, premised on the notion that digital communication would provide the necessary conditions for the world to change. In a dramatic reversal, we now associate the internet era with eroding privacy, widespread surveillance, state censorship, asymmetries of influence, and monopolies of attention—exacerbations of the exact problems it portended to fix.

Such problems are frequently understood as being problems of centralization—both infrastructural and political. If mass surveillance and censorship are problems of combined infrastructural and political centralization, then decentralization looks like a natural remedy. In the context of the internet, decentralization generally refers to peer-to-peer (p2p) technologies. In this post, I consider whether infrastructural decentralization is an effective way to counter existing regimes of political centralization. The cyber-utopian dream failed to account for the exogenous pressures that would shape the internet—the rosy narrative of infrastructural decentralization seems to be making a similar misstep.

How student debt cancellation affects the racial wealth gap

The effect of cancelling student debt on various measures of individual and group-level inequality has been a matter of controversy, especially given presidential candidates’ recent and high-profile proposals to eliminate outstanding student debt. In this work, I attempt to shed light on the policy counterfactual by analyzing the Survey of Consumer Finances for 2016, the most recent nationally-representative dataset that gives a picture of the demographics of student debt.

When we test the effects of cancelling student debt on the racial wealth gap, we conclude that across all samples, across all quantiles, the racial wealth gap narrows when student debt is cancelled, and it narrows more the more student debt is cancelled.

With respect to the two presidential candidates’ plans, this means that the Sanders plan, completely eliminating outstanding student debt, reduces racial wealth inequality more than does the Warren plan, which only forgives $50,000 of debt, and phases that out for high earners. But the difference between the two plans as measured by the reduction in the racial wealth gap is not large. It would be fair to say that the Warren plan achieves the vast majority of the racial wealth equity gains that the Sanders plan achieves, while leaving the student debt held by the highest-income borrowers intact.

A new book by James Crotty reexamines the career of John Maynard Keynes

What drives economic growth and stagnation? What types of methodologies and tools do we need to accurately explain economic epochs in the past and present? What models and policy approaches can lead to prosperity for all? These questions occupied the mind of John Maynard Keynes from World War One until his death in 1946. Keynes, one of the most influential economists of all time, is often claimed to have “saved capitalism.” His legacy, as understood by most of the economics profession, was to cure laissez-faire capitalism with countercyclical fiscal policy—using expansionary government spending during recessions to increase output and employment.

In his new book, Keynes Against Capitalism, economist James Crotty argues that this interpretation of Keynes is profoundly mistaken. Keynes, Crotty argues, wanted to replace capitalism with his own program of “liberal socialism.” Through the book, he demonstrates that 1) Keynes fundamentally rejected the theoretical model that undergirds laissez-faire capitalism; and 2) the cornerstone of Keynes’ liberal socialism program was permanent, large-scale public and semi-public investment guided by the state, accompanied by low interest rates and capital controls.

The uses of algorithms discussed in the first part of this article vary widely: from hiring decisions to bail assignment, to political campaigns and military intelligence.

Across all these applications of machine learning methods, there is a common thread: Data on individuals is used to treat different individuals differently. In the past, broadly speaking, such commercial and government activities used to target everyone in a given population more or less similarly—the same advertisements, the same prices, the same political slogans. More and more now, everyone gets personalized advertisements, personalized prices, and personalized political messages. New inequalities are created and new fragmentations of discourse are introduced.

Is that a problem? Well, it depends. I will discuss two types of concerns. The first type, relevant in particular to news and political messaging, is that the differentiation of messages is by itself a source of problems.

Terminology like "machine learning," "artificial intelligence," "deep learning," and "neural nets" is pervasive: business, universities, intelligence agencies, and political parties are all anxious to maintain an edge over the use of these technologies. Statisticians might be forgiven for thinking that this hype simply reflects the success of the marketing speak of Silicon Valley entrepreneurs vying for venture capital. All these fancy new terms are just describing something statisticians have been doing for at least two centuries.

But recent years have indeed seen impressive new achievements for various prediction problems, which are finding applications in ever more consequential aspects of society: advertising, incarceration, insurance, and war are all increasingly defined by the capacity for statistical prediction. And there is crucial a thread that ties these widely disparate applications of machine learning together: the use of data on individuals to treat different individuals differently. In this two part post, Max Kasy surveys the politics of the machine learning landscape.