An official website of the United States government
Deep neural networks (DNNs) are increasingly used in real-world applications (e.g. facial recognition). This has resulted in concerns about the fairness of decisions made by these models. Various notions and measures of fairness have been proposed to ensure that a decision-making system does not disproportionately harm (or benefit) particular subgroups of the population. In this paper, we argue that traditional notions of fairness that are only based on models' outputs are not sufficient when the model is vulnerable to adversarial attacks. We argue that in some cases, it may be easier for an attacker to target a particular subgroup, resulting in a form of robustness bias. We show that measuring robustness bias is a challenging task for DNNs and propose two methods to measure this form of bias. We then conduct an empirical study on state-of-the-art neural networks on commonly used real-world datasets such as CIFAR-10, CIFAR-100, Adience, and UTKFace and show that in almost all cases there are subgroups (in some cases based on sensitive attributes like race, gender, etc) which are less robust and are thus at a disadvantage. We argue that this kind of bias arises due to both the data distribution and the highly complex nature of the learned decision boundary in the case of DNNs, thus making mitigation of such biases a non-trivial task. Our results show that robustness bias is an important criterion to consider while auditing real-world systems that rely on DNNs for decision making. Code to reproduce all our results can be found here: https://github.com/nvedant07/Fairness-Through-Robustness
more »
« less
This content will become publicly available on October 15, 2026
A Principled Approach for Data Bias Mitigation
The widespread use of machine learning and data-driven algorithms for decision making has been steadily increasing over many years. Bias in the data can adversely affect this decision-making. We present a new mitigation strategy to address data bias. Our methods are explainable and come with mathematical guarantees of correctness. They can take advantage of new work on table discovery to find new tuples that can be added to a dataset to create real datasets that are unbiased or less biased. Our framework covers data with non-binary labels and with multiple sensitive attributes. Hence, we are able to measure and mitigate bias that does not appear over a single attribute (or feature), but only intersectionally, when considering a combination of attributes. We evaluate our techniques on publicly available datasets and provide a theoretical analysis of our results, highlighting novel insights into data bias.
more »
« less
- Award ID(s):
- 2325632
- PAR ID:
- 10647808
- Publisher / Repository:
- AAAI
- Date Published:
- Journal Name:
- Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society
- Volume:
- 8
- Issue:
- 3
- ISSN:
- 3065-8365
- Page Range / eLocation ID:
- 2273 to 2283
- Subject(s) / Keyword(s):
- Version Management Bias Bias Mitigation
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
As reliance on Machine Learning (ML) systems in real-world decision-making processes grows, ensuring these systems are free of bias against sensitive demographic groups is of increasing importance. Existing techniques for automatically debiasing ML models generally require access to either the models’ internal architectures, the models’ training datasets, or both. In this paper we outline the reasons why such requirements are disadvantageous, and present an alternative novel debiasing system that is both data- and model-agnostic. We implement this system as a Reinforcement Learning Agent and through extensive experiments show that we can debias a variety of target ML model architectures over three benchmark datasets. Our results show performance comparable to data- and/or model-gnostic state-of-the-art debiasers.more » « less
-
Machine learning (ML) is playing an increasing role in decision-making tasks that directly affect individuals, e.g., loan approvals, or job applicant screening. Significant concerns arise that, without special provisions, individuals from under-privileged backgrounds may not get equitable access to services and opportunities. Existing research studies {\em fairness} with respect to protected attributes such as gender, race or income, but the impact of location data on fairness has been largely overlooked. With the widespread adoption of mobile apps, geospatial attributes are increasingly used in ML, and their potential to introduce unfair bias is significant, given their high correlation with protected attributes. We propose techniques to mitigate location bias in machine learning. Specifically, we consider the issue of miscalibration when dealing with geospatial attributes. We focus on {\em spatial group fairness} and we propose a spatial indexing algorithm that accounts for fairness. Our KD-tree inspired approach significantly improves fairness while maintaining high learning accuracy, as shown by extensive experimental results on real data.more » « less
-
Abstract. In the geosciences, recent attention has been paid to the influence of uncertainty on expert decision-making. When making decisions under conditions of uncertainty, people tend to employ heuristics (rules of thumb) based on experience, relying on their prior knowledge and beliefs to intuitively guide choice. Over 50 years of decision-making research in cognitive psychology demonstrates that heuristics can lead to less-than-optimal decisions, collectively referred to as biases. For example, the availability bias occurs when people make judgments based on what is most dominant or accessible in memory; geoscientists who have spent the past several months studying strike-slip faults will have this terrain most readily available in their mind when interpreting new seismic data. Given the important social and commercial implications of many geoscience decisions, there is a need to develop effective interventions for removing or mitigating decision bias. In this paper, we outline the key insights from decision-making research about how to reduce bias and review the literature on debiasing strategies. First, we define an optimal decision, since improving decision-making requires having a standard to work towards. Next, we discuss the cognitive mechanisms underlying decision biases and describe three biases that have been shown to influence geoscientists' decision-making (availability bias, framing bias, anchoring bias). Finally, we review existing debiasing strategies that have applicability in the geosciences, with special attention given to strategies that make use of information technology and artificial intelligence (AI). We present two case studies illustrating different applications of intelligent systems for the debiasing of geoscientific decision-making, wherein debiased decision-making is an emergent property of the coordinated and integrated processing of human–AI collaborative teams.more » « less
-
Networks are a natural way of thinking about many datasets. The data on which a network is based, however, is rarely collected in a form that suits the analysis process, making it necessary to create and reshape networks. Data wrangling is widely acknowledged to be a critical part of the data analysis pipeline, yet interactive network wrangling has received little attention in the visualization research community. In this paper, we discuss a set of operations that are important for wrangling network datasets and introduce a visual data wrangling tool, Origraph, that enables analysts to apply these operations to their datasets. Key operations include creating a network from source data such as tables, reshaping a network by introducing new node or edge classes, filtering nodes or edges, and deriving new node or edge attributes. Our tool, Origraph, enables analysts to execute these operations with little to no programming, and to immediately visualize the results. Origraph provides views to investigate the network model, a sample of the network, and node and edge attributes. In addition, we introduce interfaces designed to aid analysts in specifying arguments for sensible network wrangling operations. We demonstrate the usefulness of Origraph in two Use Cases: first, we investigate gender bias in the film industry, and then the influence of money on the political support for the war in Yemen.more » « less
