Alexandra Chouldechova

In recent years, colleges and universities have begun to use machine learning (ML) systems to inform admissions decisions. Meanwhile, in the 2023 case Students for Fair Admissions, Inc. v. President and Fellows of Harvard College , the Supreme Court held that colleges and universities may not make admissions decisions “on the basis of race.” These parallel developments—the rise of ML in admissions and the fall of race-based affirmative action—will force educational institutions, and ultimately courts, to confront the difficult question of what it means for ML systems to differentiate “on the basis of race.” We begin by mapping the Students for Fair Admissions decision onto different uses of race in predictive AI. We distinguish between “first-order” and “second-order” race consciousness at both the training and predictive phases of machine learning, and we argue that each category of race consciousness raises distinct legal and normative issues. We go on to show that the Students for Fair Admissions decision potentially permits—and even endorses—certain forms of race consciousness. Our analysis is grounded in the observation that the process of developing ML-based systems enables policymakers to calibrate decision making algorithms much more precisely and explicitly in response to specific criticisms of race-conscious affirmative action.

Predictive risk models in the public sector are commonly developed using administrative data that is more complete for subpopulations that more greatly rely on public services. In the United States, for instance, information on health care utilization is routinely available to government agencies for individuals supported by Medicaid and Medicare, but not for the privately insured. Critiques of public sector algorithms have identified such “differential feature under-reporting” as a driver of disparities in algorithmic decision-making. Yet this form of data bias remains understudied from a technical viewpoint. While prior work has examined the fairness impacts of additive feature noise and features that are clearly marked as missing, little is known about the setting of data missingness absent indicators (i.e. differential feature under-reporting). In this work, we study an analytically tractable model of differential feature under-reporting to characterizethe impact of under-report on algorithmic fairness. We demonstrate how standard missing data methods typically fail to mitigate bias in this setting, and propose a new set of augmented loss and imputation methods. Our results show that, in real world data settings, under-reporting typically exacerbates disparities. The proposed solution methods show some success in mitigating disparities attributable to feature under-reporting.

Disaggregated evaluation is a central task in AI fairness assessment, where the goal is to measure an AI system’s performance across different subgroups defined by combinations of demographic or other sensitive attributes. The standard approach is to stratify the evaluation data across subgroups and compute performance metrics separately for each group. However, even for moderately-sized evaluation datasets, sample sizes quickly get small once considering intersectional subgroups, which greatly limits the extent to which intersectional groups are included in analysis. In this work, we introduce a structured regression approach to disaggregated evaluation that we demonstrate can yield reliable system performance estimates even for very small subgroups. We provide corresponding inference strategies for constructing confidence intervals and explore how goodness-of-fit testing can yield insight into the structure of fairness-related harms experienced by intersectional groups. We evaluate our approach on two publicly available datasets, and several variants of semi-synthetic data. The results show that our method is considerably more accurate than the standard approach, especially for small subgroups, and demonstrate how goodness-of-fit testing helps identify the key factors that drive differences in performance.

Although much literature has established the presence of demographic bias in natural language processing (NLP) models, most work relies on curated bias metrics that may not be reflective of real-world applications. At the same time, practitioners are increasingly using algorithmic tools in high-stakes settings, with particular recent interest in NLP. In this work, we focus on one such setting: child protective services (CPS). CPS workers often write copious free-form text notes about families they are working with, and CPS agencies are actively seeking to deploy NLP models to leverage these data. Given well-established racial bias in this setting, we investigate possible ways deployed NLP is liable to increase racial disparities. We specifically examine word statistics within notes and algorithmic fairness in risk prediction, coreference resolution, and named entity recognition (NER). We document consistent algorithmic unfairness in NER models, possible algorithmic unfairness in coreference resolution models, and little evidence of exacerbated racial bias in risk prediction. While there is existing pronounced criticism of risk prediction, our results expose previously undocumented risks of racial bias in realistic information extraction systems, highlighting potential concerns in deploying them, even though they may appear more benign. Our work serves as a rare realistic examination of NLP algorithmic fairness in a potential deployed setting and a timely investigation of a specific risk associated with deploying NLP in CPS settings.

Prediction models have been widely adopted as the basis for decision-making in domains as diverse as employment, education, lending, and health. Yet, few real world problems readily present themselves as precisely formulated prediction tasks. In particular, there are often many reasonable target variable options. Prior work has argued that this is an important and sometimes underappreciated choice, and has also shown that target choice can have a significant impact on the fairness of the resulting model. However, the existing literature does not offer a formal framework for characterizing the extent to which target choice matters in a particular task. Our work fills this gap by drawing connections between the problem of target choice and recent work on predictive multiplicity. Specifically, we introduce a conceptual and computational framework for assessing how the choice of target affects individuals’ outcomes and selection rate disparities across groups. We call this multi-target multiplicity. Along the way, we refine the study of single-target multiplicity by introducing notions of multiplicity that respect resource constraints—a feature of many real-world tasks that isn’t captured by existing notions of predictive multiplicity. We apply our methods on a healthcare dataset, and show that the level of multiplicity that stems from target variable choice can be greater than that stemming from nearly-optimal models of a single target.

Algorithm aversion occurs when humans are reluctant to use algorithms despite their superior performance. Studies show that giving users outcome control by providing agency over how models’ predictions are incorporated into decision-making mitigates algorithm aversion. We study whether algorithm aversion is mitigated by process control, wherein users can decide what input factors and algorithms to use in model training. We conduct a replication study of outcome control, and test novel process control study conditions on Amazon Mechanical Turk (MTurk) and Prolific. Our results partly confirm prior findings on the mitigating effects of outcome control, while also forefronting reproducibility challenges. We find that process control in the form of choosing the training algorithm mitigates algorithm aversion, but changing inputs does not. Furthermore, giving users both outcome and process control does not reduce algorithm aversion more than outcome or process control alone. This study contributes to design considerations around mitigating algorithm aversion.