Should Fairness be a Metric or a Model? A Model-based Framework for Assessing Bias in Machine Learning Pipelines

Abstract

Fairness measurement is crucial for assessing algorithmic bias in various types of machine learning (ML) models, including ones used for search relevance, recommendation, personalization, talent analytics, and natural language processing. However, the fairness measurement paradigm is currently dominated by fairness metrics that examine disparities in allocation and/or prediction error as univariate key performance indicators (KPIs) for a protected attribute or group. Although important and effective in assessing ML bias in certain contexts such as recidivism, existing metrics don’t work well in many real-world applications of ML characterized by imperfect models applied to an array of instances encompassing a multivariate mixture of protected attributes, that are part of a broader process pipeline. Consequently, the upstream representational harm quantified by existing metrics based on how the model represents protected groups doesn’t necessarily relate to allocational harm in the application of such models in downstream policy/decision contexts. We propose FAIR-Frame, a model-based framework for parsimoniously modeling fairness across multiple protected attributes in regard to the representational and allocational harm associated with the upstream design/development and downstream usage of ML models. We evaluate the efficacy of our proposed framework on two testbeds pertaining to text classification using pretrained language models. The upstream testbeds encompass over fifty thousand documents associated with twenty-eight thousand users, seven protected attributes and five different classification tasks. The downstream testbeds span three policy outcomes and over 5.41 million total observations. Results in comparison with several existing metrics show that the upstream representational harm measures produced by FAIR-Frame and other metrics are significantly different from one another, and that FAIR-Frame’s representational fairness measures have the highest percentage alignment and lowest error with allocational harm observed in downstream applications. Our findings have important implications for various ML contexts, including information retrieval, user modeling, digital platforms, and text classification, where responsible and trustworthy AI are becoming an imperative.

Learning-Based Models for Building User Profiles for Personalized Information Access

“…To address concerns about algorithmic fairness and bias in personalized recommendation systems, recent research has focused on developing fairness-aware models and evaluation metrics (Gao et al, 2022;Lalor et al, 2024;Zhang et al, 2023). These studies aim to mitigate bias and ensure equitable treatment across diverse user groups, promoting diversity and inclusivity in personalized recommendations.…”

A study of the motivation of clerical officers in the Hong Kong Polytechnic University

“…Contrary to the implicit assumptions underlying various formal models of information seeking, it is posited that users exhibit bounded rationality, and typically do not base their search decisions on precise estimations of search gains and costs [2,55]. Despite the increasing attention paid to data biases and algorithm biases in the the field of computing [13,44,47], only a few studies have focused on the issue of properly recognizing and effectively mitigating the negative effects of biases on human decision-making processes. Within the realm of cognitive biases influencing decision-making, the decoy effect reflects how users alter their in-situ preferences and judgments on presented options.…”