Kearns, Neel, Roth, and Wu [ICML 2018] recently proposed a notion
of rich subgroup fairness intended to bridge the gap between statistical and individual notions of fairness. Rich subgroup fairness picks
a statistical fairness constraint (say, equalizing false positive rates
across protected groups), but then asks that this constraint hold
over an exponentially or infinitely large collection of subgroups defined by a class of functions with bounded VC dimension. They give
an algorithm guaranteed to learn subject to this constraint, under
the condition that it has access to oracles for perfectly learning absent a fairness constraint. In this paper, we undertake an extensive
empirical evaluation of the algorithm of Kearns et al. On four real
datasets for which fairness is a concern, we investigate the basic
convergence of the algorithm when instantiated with fast heuristics
in place of learning oracles, measure the tradeoffs between fairness
and accuracy, and compare this approach with the recent algorithm
of Agarwal, Beygelzeimer, Dudik, Langford, and Wallach [ICML
2018], which implements weaker and more traditional marginal
fairness constraints defined by individual protected attributes. We
find that in general, the Kearns et al. algorithm converges quickly,
large gains in fairness can be obtained with mild costs to accuracy,
and that optimizing accuracy subject only to marginal fairness
leads to classifiers with substantial subgroup unfairness. We also
provide a number of analyses and visualizations of the dynamics
and behavior of the Kearns et al. algorithm. Overall we find this
algorithm to be effective on real data, and rich subgroup fairness to
be a viable notion in practice
more »
« less
An Algorithmic Framework for Fairness Elicitation
We consider settings in which the right notion of fairness is not captured by simple mathematical definitions (such as equality of error rates across groups), but might be more complex and nuanced and thus require elicitation from individual or collective stakeholders. We introduce a framework in which pairs of individuals can be identified as requiring (approximately) equal treatment under a learned model, or requiring ordered treatment such as "applicant Alice should be at least as likely to receive a loan as applicant Bob". We provide a provably convergent and oracle efficient algorithm for learning the most accurate model subject to the elicited fairness constraints, and prove generalization bounds for both accuracy and fairness. This algorithm can also combine the elicited constraints with traditional statistical fairness notions, thus "correcting" or modifying the latter by the former. We report preliminary findings of a behavioral study of our framework using human-subject fairness constraints elicited on the COMPAS criminal recidivism dataset.
more »
« less
- NSF-PAR ID:
- 10267302
- Date Published:
- Journal Name:
- 2nd Symposium on Foundations of Responsible Computing (FORC 2021)
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Recently, there has been a growing interest in developing machine learning (ML) models that can promote fairness, i.e., eliminating biased predictions towards certain populations (e.g., individuals from a specific demographic group). Most existing works learn such models based on well-designed fairness constraints in optimization. Nevertheless, in many practical ML tasks, only very few labeled data samples can be collected, which can lead to inferior fairness performance. This is because existing fairness constraints are designed to restrict the prediction disparity among different sensitive groups, but with few samples, it becomes difficult to accurately measure the disparity, thus rendering ineffective fairness optimization. In this paper, we define the fairness-aware learning task with limited training samples as the fair few-shot learning problem. To deal with this problem, we devise a novel framework that accumulates fairness-aware knowledge across different meta-training tasks and then generalizes the learned knowledge to meta-test tasks. To compensate for insufficient training samples, we propose an essential strategy to select and leverage an auxiliary set for each meta-test task. These auxiliary sets contain several labeled training samples that can enhance the model performance regarding fairness in meta-test tasks, thereby allowing for the transfer of learned useful fairness-oriented knowledge to meta-test tasks. Furthermore, we conduct extensive experiments on three real-world datasets to validate the superiority of our framework against the state-of-the-art baselines.more » « less
-
This Article develops a framework for both assessing and designing content moderation systems consistent with public values. It argues that moderation should not be understood as a single function, but as a set of subfunctions common to all content governance regimes. By identifying the particular values implicated by each of these subfunctions, it explores the appropriate ways the constituent tasks might best be allocated-specifically to which actors (public or private, human or technological) they might be assigned, and what constraints or processes might be required in their performance. This analysis can facilitate the evaluation and design of content moderation systems to ensure the capacity and competencies necessary for legitimate, distributed systems of content governance. Through a combination of methods, legal schemes delegate at least a portion of the responsibility for governing online expression to private actors. Sometimes, statutory schemes assign regulatory tasks explicitly. In others, this delegation often occurs implicitly, with little guidance as to how the treatment of content should be structured. In the law's shadow, online platforms are largely given free rein to configure the governance of expression. Legal scholarship has surfaced important concerns about the private sector's role in content governance. In response, private platforms engaged in content moderation have adopted structures that mimic public governance forms. Yet, we largely lack the means to measure whether these forms are substantive, effectively infusing public values into the content moderation process, or merely symbolic artifice designed to deflect much needed public scrutiny. This Article's proposed framework addresses that gap in two ways. First, the framework considers together all manner of legal regimes that induce platforms to engage in the function of content moderation. Second, it focuses on the shared set of specific tasks, or subfunctions, involved in the content moderation function across these regimes. Examining a broad range of content moderation regimes together highlights the existence of distinct common tasks and decision points that together constitute the content moderation function. Focusing on this shared set of subfunctions highlights the different values implicated by each and the way they can be "handed off' to human and technical actors to perform in different ways with varying normative and political implications. This Article identifies four key content moderation subfunctions: (1) definition of policies, (2) identification of potentially covered content, (3) application of policies to specific cases, and (4) resolution of those cases. Using these four subfunctions supports a rigorous analysis of how to leverage the capacities and competencies of government and private parties throughout the content moderation process. Such attention also highlights how the exercise of that power can be constrained-either by requiring the use of particular decision-making processes or through limits on the use of automation-in ways that further address normative concerns. Dissecting the allocation of subfunctions in various content moderation regimes reveals the distinct ethical and political questions that arise in alternate configurations. Specifically, it offers a way to think about four key questions: (1) what values are most at issue regarding each subfunction; (2) which activities might be more appropriate to delegate to particular public or private actors; (3) which constraints must be attached to the delegation of each subfunction; and (4) where can investments in shared content moderation infrastructures support relevant values? The functional framework thus provides a means for both evaluating the symbolic legal forms that firms have constructed in service of content moderation and for designing processes that better reflect public values.more » « less
-
Motivated by settings in which predictive models may be required to be non-discriminatory with respect to certain attributes (such as race), but even collecting the sensitive attribute may be forbidden or restricted, we initiate the study of fair learning under the constraint of differential privacy. Our first algorithm is a private implementation of the equalized odds post-processing approach of (Hardt et al., 2016). This algorithm is appealingly simple, but must be able to use protected group membership explicitly at test time, which can be viewed as a form of “disparate treatment”. Our second algorithm is a differentially private version of the oracle-efficient in-processing approach of (Agarwal et al., 2018) which is more complex but need not have access to protected group membership at test time. We identify new tradeoffs between fairness, accuracy, and privacy that emerge only when requiring all three properties, and show that these tradeoffs can be milder if group membership may be used at test time. We conclude with a brief experimental evaluation.more » « less
-
While implicit feedback (e.g., clicks, dwell times, etc.) is an abundant and attractive source of data for learning to rank, it can produce unfair ranking policies for both exogenous and endogenous reasons. Exogenous reasons typically manifest themselves as biases in the training data, which then get reflected in the learned ranking policy and often lead to rich-get-richer dynamics. Moreover, even after the correction of such biases, reasons endogenous to the design of the learning algorithm can still lead to ranking policies that do not allocate exposure among items in a fair way. To address both exogenous and endogenous sources of unfairness, we present the first learning-to-rank approach that addresses both presentation bias and merit-based fairness of exposure simultaneously. Specifically, we define a class of amortized fairness-of-exposure constraints that can be chosen based on the needs of an application, and we show how these fairness criteria can be enforced despite the selection biases in implicit feedback data. The key result is an efficient and flexible policy-gradient algorithm, called FULTR, which is the first to enable the use of counterfactual estimators for both utility estimation and fairness constraints. Beyond the theoretical justification of the framework, we show empirically that the proposed algorithm can learn accurate and fair ranking policies from biased and noisy feedback.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.4230/LIPIcs.FORC.2021.2
