Recommender systems are information filtering tools that discover potential matchings between users and items and benefit both parties. This benefit can be considered a social resource that should be equitably allocated across users and items, especially in critical domains such as education and employment. Biases and unfairness in recommendations raise both ethical and legal concerns. In this dissertation, we investigate the concept of unfairness in the context of recommender systems. In particular, we study appropriate unfairness evaluation metrics, examine the relation between bias in recommender models and inequality in the underlying population, as well as propose effective unfairness mitigation approaches.
We start with exploring the implication of fairness in recommendation and formulating unfairness evaluation metrics. We focus on the task of rating prediction. We identify the insufficiency of demographic parity for scenarios where the target variable is justifiably dependent on demographic features. Then we propose an alternative set of unfairness metrics that measured based on how much the average predicted ratings deviate from average true ratings. We also reduce these unfairness in matrix factorization (MF) models by explicitly adding them as penalty terms to learning objectives.
Next, we target a form of unfairness in matrix factorization models observed as disparate model performance across user groups. We identify four types of biases in the training data that contribute to higher subpopulation error. Then we propose personalized regularization learning (PRL), which learns personalized regularization parameters that directly address the data biases. PRL poses the hyperparameter search problem as a secondary learning task. It enables back-propagation to learn the personalized regularization parameters by leveraging the closed-form solutions of alternating least squares (ALS) to solve MF. Furthermore, the learned parameters are interpretable and provide insights into how fairness is improved.
Third, we conduct theoretical analysis on the long-term dynamics of inequality in the underlying population, in terms of the fitting between users and items. We view the task of recommendation as solving a set of classification problems through threshold policies. We mathematically formulate the transition dynamics of user-item fit in one step of recommendation. Then we prove that a system with the formulated dynamics always has at least one equilibrium, and we provide sufficient conditions for the equilibrium to be unique. We also show that, depending on the item category relationships and the recommendation policies, recommendations in one item category can reshape the user-item fit in another item category.
To summarize, in this research, we examine different fairness criteria in rating prediction and recommendation, study the dynamic of interactions between recommender systems and users, and propose mitigation methods to promote fairness and equality. / Doctor of Philosophy / Recommender systems are information filtering tools that discover potential matching between users and items. However, a recommender system, if not properly built, may not treat users and items equitably, which raises ethical and legal concerns. In this research, we explore the implication of fairness in the context of recommender systems, study the relation between unfairness in recommender output and inequality in the underlying population, and propose effective unfairness mitigation approaches.
We start with finding unfairness metrics appropriate for recommender systems. We focus on the task of rating prediction, which is a crucial step in recommender systems. We propose a set of unfairness metrics measured as the disparity in how much predictions deviate from the ground truth ratings. We also offer a mitigation method to reduce these forms of unfairness in matrix factorization models
Next, we look deeper into the factors that contribute to error-based unfairness in matrix factorization models and identify four types of biases that contribute to higher subpopulation error. Then we propose personalized regularization learning (PRL), which is a mitigation strategy that learns personalized regularization parameters to directly addresses data biases. The learned per-user regularization parameters are interpretable and provide insight into how fairness is improved.
Third, we conduct a theoretical study on the long-term dynamics of the inequality in the fitting (e.g., interest, qualification, etc.) between users and items. We first mathematically formulate the transition dynamics of user-item fit in one step of recommendation. Then we discuss the existence and uniqueness of system equilibrium as the one-step dynamics repeat. We also show that depending on the relation between item categories and the recommendation policies (unconstrained or fair), recommendations in one item category can reshape the user-item fit in another item category.
In summary, we examine different fairness criteria in rating prediction and recommendation, study the dynamics of interactions between recommender systems and users, and propose mitigation methods to promote fairness and equality.
| Identifer | oai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/103779 |
| Date | 10 June 2021 |
| Creators | Yao, Sirui |
| Contributors | Computer Science, Huang, Bert, Ramakrishnan, Narendran, Beutel, Alex, Prakash, B. Aditya, Reddy, Chandan K. |
| Publisher | Virginia Tech |
| Source Sets | Virginia Tech Theses and Dissertation |
| Detected Language | English |
| Type | Dissertation |
| Format | ETD, application/pdf |
| Rights | In Copyright, http://rightsstatements.org/vocab/InC/1.0/ |
Page generated in 0.002 seconds