This paper explores the problem of “underspecification” – a statistical phenomenon that occurs when an observed issue may have several possible causes, not all of which are accounted for in the model. The team of authors from Google examined case studies in computer vision, medical imaging, natural language processing, and medical genomics, and found variation in model performance based on underspecification problems using a variety of ML pipelines. As a result, training processes that can produce sound models often result in poor models, and the difference between the two will not be apparent until the model is in use and has to generalize to non-training data. Based on these findings, the authors point to greater rigor in specifying model requirements and stress testing models before they are approved for use.

The authors explore the application of modern antidiscrimination law to algorithmic fairness techniques and find incompatibility between those approaches and equal protection jurisprudence that demands “individualized consideration” and bars formal, quantitative weights for race regardless of purpose. The authors look to government-contracting cases as an alternative grounding for algorithmic fairness, because these cases permit explicit and quantitative race-based remedies based on historical discrimination by the actor. But while limited, this doctrinal approach mandates that adjustments be calibrated to the entity’s responsibility for historical discrimination causing present-day disparities. The authors argue that these cases provide a legally viable path for algorithmic fairness under current constitutional doctrine but call for more research at the intersection of algorithmic fairness and causal inference to ensure that bias mitigation is tailored to specific causes and mechanisms of bias.

The author considers the complexity of using algorithmic decision-making in policy-sensitive areas, like determining criminal bail and sentences and welfare benefits claims and argues that advances in explainability techniques are necessary, but not sufficient, for resolving key questions about such decisions. She argues that the inherent complexity of the most powerful AI models and our inability to reduce law and regulation to clearly stated optimization goals for the algorithm reinforce the need for transparent governance by model users, especially when they are government agencies..

This article analyzes AI fairness as both essential in itself and as a way to solve the issue of trust in AI systems. The author advocates for an interdisciplinary approach, with computer science and the social sciences working together. Three recommendations are outlined: (1) train managers to act as “devil’s advocates” by evaluating algorithmic decision-making using common sense and intuitive notions of what is right and wrong; (2) require leaders to articulate their companies’ values and moral norms to help inform compromises between utility and human values in AI deployment; (3) hold data scientists and organizational leaders responsible for collaborating to evaluate the fairness of AI models both against technical definitions and broader company values.

This source collects recent trends in short-term forbearances in the mortgage market but also notes areas in which additional data and consumer outreach are urgently needed. In particular, it highlights that about 530,000 homeowners who became delinquent after the pandemic did not take advantage of forbearance, despite being eligible to ask for relief under federal legislation. An additional 205,000 homeowners obtained an initial forbearance that expired in June or July, but did not seek to extended it and have since become delinquent.

This study of loan-level Paycheck Protection Program data finds that despite a lag in approving several fintech lenders to participate in the program, such lenders provided disproportionate amounts of PPP funds in ZIP codes with fewer bank branches, lower incomes, and a larger minority share of the population, as well as in industries with little ex ante small-business lending. Fintechs’ role in PPP provision was also greater in counties where the economic effects of the COVID-19 pandemic were more severe.

This paper analyzes weekly eviction filings for 44 jurisdictions in 11 states through July 7 compared to the same period in 2019. It finds that filings have returned to pre-pandemic levels in jurisdictions that were not subject to restrictions, and that activity has surged in jurisdictions that prohibited both filings and hearings immediately after the pandemic.​

This paper provides the first analysis of impacts of the pandemic on the number of active small businesses in the United States using nationally representative data from May 2020. The number of active business owners rebounded 7% since the low in April but remained 15% down from February. Drops in business activity from pre-pandemic levels are disproportionately concentrated among African-American (26%), Latino (19%), Asian (21%), and immigrant (25%) business owners.

This paper analyzes cell phone data from March through mid May, finding that private, self-regulating changes in behavior explained about 75% of the decline in foot traffic across most industries, while restrictive regulation (including school closures) had more influence on essential retail foot traffic and the fraction of cell phones that remained at home all day.

This article explores how Wall Street is mining geolocational data, payments information, social media posts, and various other sources of data in an effort to better understand emerging trends from the pandemic. While useful in various ways, it also cautions about the risk of unrepresentative data.