Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification (2019-03-11T00:00:00.000000Z)

TL;DR

A suite of threshold-agnostic metrics are introduced that provide a nuanced view of unintended bias in Machine Learning, by considering the various ways that a classifier’s score distribution can vary across designated groups.

Abstract

Unintended bias in Machine Learning can manifest as systemic differences in performance for different demographic groups, potentially compounding existing challenges to fairness in society at large. In this paper, we introduce a suite of threshold-agnostic metrics that provide a nuanced view of this unintended bias, by considering the various ways that a classifier’s score distribution can vary across designated groups. We also introduce a large new test set of online comments with crowd-sourced annotations for identity references. We use this to show how our metrics can be used to find new and potentially subtle unintended bias in existing public models.

Authors

Nithum Thain

3 papers

Lucas Dixon

2 papers

Daniel Borkan

1 papers

Nuanced Metrics for Measuring Unintended Bias with Real Data for Text Classification

TL;DR

Abstract

Authors

References20 items

The (Im)possibility of fairness

Limitations of Pinned AUC for Measuring Unintended Bias

Measuring and Mitigating Unintended Bias in Text Classification

Model Cards for Model Reporting

Counterfactual Fairness in Text Classification through Robustness

Reducing Gender Bias in Abusive Language Detection

Adversarial Removal of Demographic Attributes from Text Data

Gender Shades: Intersectional Accuracy Disparities in Commercial Gender Classification

The cost of fairness in binary classification

Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations

Wikipedia Talk Labels: Toxicity

Ex Machina: Personal Attacks Seen at Scale

Equality of Opportunity in Supervised Learning

Inherent Trade-Offs in the Fair Determination of Risk Scores

Certifying and Removing Disparate Impact

Evaluating the Predictive Validity of the Compas Risk and Needs Assessment System

Areas beneath the relative operating characteristics (ROC) and relative operating levels (ROL) curves: Statistical significance and interpretation

Unintended bias and names of frequently targeted groups

Perspective API

Background Positive Subgroup Negative (BPSN) AUC Equation term 2 calculates AUC

Field of Study

Journal Information

Name

Venue Information

Name

Type

URL

Alternate Names