Models trained by means of supervised learning are increasingly deployed in high-stakes domains, and, when their predictions inform decisions about people, they inevitably impact (positively or negatively) on their lives. As a consequence, those in charge of developing these models must carefully evaluate their impact on different groups of people and ensure that sensitive demographic attributes, such as race or sex, do not result in unfair treatment for members of specific groups. For doing this, awareness of demographic attributes on the part of those evaluating model impacts is fundamental. Unfortunately, the collection of these attributes is often in conflict with industry practices and legislation on data minimization and privacy. For this reason, it may be hard to measure the group fairness of trained models, even from within the companies developing them. In this work, we tackle the problem of measuring group fairness under unawareness of sensitive attributes, by using techniques from quantification, a supervised learning task concerned with directly providing group-level prevalence estimates (rather than individual-level class labels). We identify five important factors that complicate the estimation of fairness under unawareness and formalize them into five different experimental protocols under which we assess the effectiveness of different estimators of group fairness. We also consider the problem of potential model misuse to infer sensitive attributes at an individual level, and demonstrate that quantification approaches are suitable for decoupling the (desirable) objective of measuring group fairness from the (undesirable) objective of inferring sensitive attributes of individuals.
相關內容
Federated learning (FL) has emerged as an important machine learning paradigm where a global model is trained based on the private data from distributed clients. However, most of existing FL algorithms cannot guarantee the performance fairness towards different clients or different groups of samples because of the distribution shift. Recent researches focus on achieving fairness among clients, but they ignore the fairness towards different groups formed by sensitive attribute(s) (e.g., gender and/or race), which is important and practical in real applications. To bridge this gap, we formulate the goal of unified group fairness on FL which is to learn a fair global model with similar performance on different groups. To achieve the unified group fairness for arbitrary sensitive attribute(s), we propose a novel FL algorithm, named Group Distributionally Robust Federated Averaging (G-DRFA), which mitigates the distribution shift across groups with theoretical analysis of convergence rate. Specifically, we treat the performance of the federated global model at each group as an objective and employ the distributionally robust techniques to maximize the performance of the worst-performing group over an uncertainty set by group reweighting. We validate the advantages of the G-DRFA algorithm with various kinds of distribution shift settings in experiments, and the results show that G-DRFA algorithm outperforms the existing fair federated learning algorithms on unified group fairness.
Background: Estimations of causal effects from observational data are subject to various sources of bias. These biases can be adjusted by using negative control outcomes not affected by the treatment. The empirical calibration procedure uses negative controls to calibrate p-values and both negative and positive controls to calibrate coverage of the 95% confidence interval of the outcome of interest. Although empirical calibration has been used in several large observational studies, there is no systematic examination of its effect under different bias scenarios. Methods: The effect of empirical calibration of confidence intervals was analyzed using simulated datasets with known treatment effects. The simulations were for binary treatment and binary outcome, with simulated biases resulting from unmeasured confounder, model misspecification, measurement error, and lack of positivity. The performance of empirical calibration was evaluated by determining the change of the confidence interval coverage and bias of the outcome of interest. Results: Empirical calibration increased coverage of the outcome of interest by the 95% confidence interval under most settings but was inconsistent in adjusting the bias of the outcome of interest. Empirical calibration was most effective when adjusting for unmeasured confounding bias. Suitable negative controls had a large impact on the adjustment made by empirical calibration, but small improvements in the coverage of the outcome of interest were also observable when using unsuitable negative controls. Conclusions: This work adds evidence to the efficacy of empirical calibration on calibrating the confidence intervals of treatment effects in observational studies. We recommend empirical calibration of confidence intervals, especially when there is a risk of unmeasured confounding.
In an ideal world, deployed machine learning models will enhance our society. We hope that those models will provide unbiased and ethical decisions that will benefit everyone. However, this is not always the case; issues arise from the data curation process to the models' deployment. The continued use of biased datasets and processes will adversely damage communities and increase the cost to fix the problem. In this work, we walk through the decision process that a researcher will need to make before, during, and after their project to consider the broader impacts of research and the community. Throughout this paper, we observe the critical decisions that are often overlooked when deploying AI, argue for the use of fairness forensics to discover bias and fairness issues in systems, assert the need for a responsible human-over-the-loop to bring accountability into the deployed system, and finally, reflect on the need to explore research agendas that have harmful societal impacts. We examine visual privacy research and draw lessons that can apply broadly to Artificial Intelligence. Our goal is to provide a systematic analysis of the machine learning pipeline for visual privacy and bias issues. With this pipeline, we hope to raise stakeholder (e.g., researchers, modelers, corporations) awareness as these issues propagate in the various machine learning phases.
Recommender systems are gaining increasing and critical impacts on human and society since a growing number of users use them for information seeking and decision making. Therefore, it is crucial to address the potential unfairness problems in recommendations. Just like users have personalized preferences on items, users' demands for fairness are also personalized in many scenarios. Therefore, it is important to provide personalized fair recommendations for users to satisfy their personalized fairness demands. Besides, previous works on fair recommendation mainly focus on association-based fairness. However, it is important to advance from associative fairness notions to causal fairness notions for assessing fairness more properly in recommender systems. Based on the above considerations, this paper focuses on achieving personalized counterfactual fairness for users in recommender systems. To this end, we introduce a framework for achieving counterfactually fair recommendations through adversary learning by generating feature-independent user embeddings for recommendation. The framework allows recommender systems to achieve personalized fairness for users while also covering non-personalized situations. Experiments on two real-world datasets with shallow and deep recommendation algorithms show that our method can generate fairer recommendations for users with a desirable recommendation performance.
AI is undergoing a paradigm shift with the rise of models (e.g., BERT, DALL-E, GPT-3) that are trained on broad data at scale and are adaptable to a wide range of downstream tasks. We call these models foundation models to underscore their critically central yet incomplete character. This report provides a thorough account of the opportunities and risks of foundation models, ranging from their capabilities (e.g., language, vision, robotics, reasoning, human interaction) and technical principles(e.g., model architectures, training procedures, data, systems, security, evaluation, theory) to their applications (e.g., law, healthcare, education) and societal impact (e.g., inequity, misuse, economic and environmental impact, legal and ethical considerations). Though foundation models are based on standard deep learning and transfer learning, their scale results in new emergent capabilities,and their effectiveness across so many tasks incentivizes homogenization. Homogenization provides powerful leverage but demands caution, as the defects of the foundation model are inherited by all the adapted models downstream. Despite the impending widespread deployment of foundation models, we currently lack a clear understanding of how they work, when they fail, and what they are even capable of due to their emergent properties. To tackle these questions, we believe much of the critical research on foundation models will require deep interdisciplinary collaboration commensurate with their fundamentally sociotechnical nature.
Algorithmic fairness has aroused considerable interests in data mining and machine learning communities recently. So far the existing research has been mostly focusing on the development of quantitative metrics to measure algorithm disparities across different protected groups, and approaches for adjusting the algorithm output to reduce such disparities. In this paper, we propose to study the problem of identification of the source of model disparities. Unlike existing interpretation methods which typically learn feature importance, we consider the causal relationships among feature variables and propose a novel framework to decompose the disparity into the sum of contributions from fairness-aware causal paths, which are paths linking the sensitive attribute and the final predictions, on the graph. We also consider the scenario when the directions on certain edges within those paths cannot be determined. Our framework is also model agnostic and applicable to a variety of quantitative disparity measures. Empirical evaluations on both synthetic and real-world data sets are provided to show that our method can provide precise and comprehensive explanations to the model disparities.
Machine learning plays a role in many deployed decision systems, often in ways that are difficult or impossible to understand by human stakeholders. Explaining, in a human-understandable way, the relationship between the input and output of machine learning models is essential to the development of trustworthy machine-learning-based systems. A burgeoning body of research seeks to define the goals and methods of explainability in machine learning. In this paper, we seek to review and categorize research on counterfactual explanations, a specific class of explanation that provides a link between what could have happened had input to a model been changed in a particular way. Modern approaches to counterfactual explainability in machine learning draw connections to the established legal doctrine in many countries, making them appealing to fielded systems in high-impact areas such as finance and healthcare. Thus, we design a rubric with desirable properties of counterfactual explanation algorithms and comprehensively evaluate all currently-proposed algorithms against that rubric. Our rubric provides easy comparison and comprehension of the advantages and disadvantages of different approaches and serves as an introduction to major research themes in this field. We also identify gaps and discuss promising research directions in the space of counterfactual explainability.
We investigate the problem of fair recommendation in the context of two-sided online platforms, comprising customers on one side and producers on the other. Traditionally, recommendation services in these platforms have focused on maximizing customer satisfaction by tailoring the results according to the personalized preferences of individual customers. However, our investigation reveals that such customer-centric design may lead to unfair distribution of exposure among the producers, which may adversely impact their well-being. On the other hand, a producer-centric design might become unfair to the customers. Thus, we consider fairness issues that span both customers and producers. Our approach involves a novel mapping of the fair recommendation problem to a constrained version of the problem of fairly allocating indivisible goods. Our proposed FairRec algorithm guarantees at least Maximin Share (MMS) of exposure for most of the producers and Envy-Free up to One item (EF1) fairness for every customer. Extensive evaluations over multiple real-world datasets show the effectiveness of FairRec in ensuring two-sided fairness while incurring a marginal loss in the overall recommendation quality.
There has been considerable growth and interest in industrial applications of machine learning (ML) in recent years. ML engineers, as a consequence, are in high demand across the industry, yet improving the efficiency of ML engineers remains a fundamental challenge. Automated machine learning (AutoML) has emerged as a way to save time and effort on repetitive tasks in ML pipelines, such as data pre-processing, feature engineering, model selection, hyperparameter optimization, and prediction result analysis. In this paper, we investigate the current state of AutoML tools aiming to automate these tasks. We conduct various evaluations of the tools on many datasets, in different data segments, to examine their performance, and compare their advantages and disadvantages on different test cases.
Rankings of people and items are at the heart of selection-making, match-making, and recommender systems, ranging from employment sites to sharing economy platforms. As ranking positions influence the amount of attention the ranked subjects receive, biases in rankings can lead to unfair distribution of opportunities and resources, such as jobs or income. This paper proposes new measures and mechanisms to quantify and mitigate unfairness from a bias inherent to all rankings, namely, the position bias, which leads to disproportionately less attention being paid to low-ranked subjects. Our approach differs from recent fair ranking approaches in two important ways. First, existing works measure unfairness at the level of subject groups while our measures capture unfairness at the level of individual subjects, and as such subsume group unfairness. Second, as no single ranking can achieve individual attention fairness, we propose a novel mechanism that achieves amortized fairness, where attention accumulated across a series of rankings is proportional to accumulated relevance. We formulate the challenge of achieving amortized individual fairness subject to constraints on ranking quality as an online optimization problem and show that it can be solved as an integer linear program. Our experimental evaluation reveals that unfair attention distribution in rankings can be substantial, and demonstrates that our method can improve individual fairness while retaining high ranking quality.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191