The Zoo of Fairness Metrics in Machine Learning

Decisions made by various artificial intelligence (AI) systems greatly influence our day-to-day lives. With the increasing use of AI systems, it becomes crucial to know that they are fair, identify the underlying biases in their decision-making, and create a standardized framework to ascertain their fairness. Biases in AI systems lead to unintended ethical, social and even legal issues. In this paper, we propose a novel and versatile Fairness Score and Bias Index for measuring the fairness of a supervised learning AI system. We also propose a standard operating procedure (SOP) for issuing Fairness Certification for such data-driven applications. Fairness Score and audit process standardization will ensure quality, reduce ambiguity, enable comparison and improve the trustworthiness of the AI systems. It will also provide a framework to operationalize the concept of fairness and facilitate the commercial deployment of such systems. Furthermore, a Fairness Certificate issued by a designated third-party auditing agency following the standardized process would boost the conviction of the organizations in the AI systems that they intend to deploy. The Bias Index proposed in this paper reveals comparative bias amongst the various protected attributes within the dataset, and the Fairness Score provides overall fairness with an ideal value of 1. To substantiate the proposed framework, we iteratively train a model on biased and unbiased data using multiple labelled datasets and check that the Fairness Score, Bias Indexes and the proposed process correctly identify the biases and judge the fairness.

show abstract

“…It can be preferred when discrimination can be accepted as long as it is justified by actual trustable data [47].…”

Section: Average Oddsmentioning

confidence: 99%

Fairness Score and process standardization: framework for fairness certification in artificial intelligence systems

Avinash¹,

Harsh²,

Agarwal

2022

AI Ethics

View full text Add to dashboard Cite

show abstract

“…Various fairness definitions and measures have been proposed and studied in the literature, such as demographic parity (group fairness, statistical parity [15]), equalized odds (disparate mistreatment), equal opportunity [23], and predictive parity [6,60]. Due to the non-differentiability of these fairness metrics, a large body of work has focused on optimizing surrogates for these metrics include directly approximating AUCPR or pairwise AUCROC loss [16], and learning the parameters of a general structured loss function [3].…”

Section: Differentiable Fairness Metricsmentioning

confidence: 99%

FORML: Learning to Reweight Data for Fairness

Yan¹,

Seto²,

Apostoloff³

2022

Preprint

View full text Add to dashboard Cite

Deployed machine learning models are evaluated by multiple metrics beyond accuracy, such as fairness and robustness. However, such models are typically trained to minimize the average loss for a single metric, which is typically a proxy for accuracy. Training to optimize a single metric leaves these models prone to fairness violations, especially when the population of sub-groups in the training data are imbalanced. This work addresses the challenge of jointly optimizing fairness and predictive performance in the multi-class classification setting by introducing Fairness Optimized Reweighting via Meta-Learning (FORML), a training algorithm that balances fairness constraints and accuracy by jointly optimizing training sample weights and a neural network's parameters. The approach increases fairness by learning to weight each training datum's contribution to the loss according to its impact on reducing fairness violations, balancing the contributions from both over-and under-represented sub-groups. We empirically validate FORML on a range of benchmark and real-world classification datasets and show that our approach improves equality of opportunity fairness criteria over existing state-of-the-art re-weighting methods by ≈ 1% on image classification tasks and by ≈ 5% on a face attribute prediction task. This improvement is achieved without pre-processing data or post-processing model outputs, without learning an additional weighting function, and while maintaining accuracy on the original predictive metric.

show abstract

“…The following is a brief introduction to constrained sequential decision-making, causal modeling [38] and fairness principles. We focus on the key essentials on fairness and refer readers desiring more detail to comprehensive surveys [44,8].…”

Section: Preliminariesmentioning

confidence: 99%

“…Numerous fairness principles have been studied across various research communities, from machine learning to economics [31,33,8,32,45,29]. Here, we categorize existing fairness principles into four main groups:…”

Section: Fairness Principlesmentioning

confidence: 99%

See 1 more Smart Citation

SCALES: From Fairness Principles to Constrained Decision-Making

Balakrishnan

Soh

2022

Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society

View full text Add to dashboard Cite

This paper proposes SCALES, a general framework that translates well-established fairness principles into a common representation based on the Constraint Markov Decision Process (CMDP). With the help of causal language, our framework can place constraints on both the procedure of decision making (procedural fairness) as well as the outcomes resulting from decisions (outcome fairness). Specifically, we show that well-known fairness principles can be encoded either as a utility component, a non-causal component, or a causal component in a SCALES-CMDP. We illustrate SCALES using a set of case studies involving a simulated healthcare scenario and the real-world COMPAS dataset. Experiments demonstrate that our framework produces fair policies that embody alternative fairness principles in single-step and sequential decision-making scenarios.

show abstract

The Zoo of Fairness Metrics in Machine Learning

Cited by 12 publications

References 43 publications

Fairness Score and process standardization: framework for fairness certification in artificial intelligence systems

Fairness Score and process standardization: framework for fairness certification in artificial intelligence systems

FORML: Learning to Reweight Data for Fairness

SCALES: From Fairness Principles to Constrained Decision-Making

Contact Info

Product

Resources

About