Date on Master's Thesis/Doctoral Dissertation


Document Type

Master's Thesis

Degree Name



Computer Engineering and Computer Science

Degree Program

Computer Science, MS

Committee Chair

Nasraoui, Olfa

Committee Member

Frigui, Hichem

Committee Member

Sanders, Scott

Author's Keywords

machine learning; recommender systems; bias; data science


Recommender systems are becoming widely used in everyday life. They use machine learning algorithms which learn to predict our preferences and thus influence our choices among a staggering array of options online, such as movies, books, products, and even news articles. Thus what we discover and see online, and consequently our opinions and decisions, are becoming increasingly affected by automated predictions made by learning machines. Similarly, the predictive accuracy of these learning machines heavily depends on the feedback data, such as ratings and clicks, that we provide them. This mutual influence can lead to closed-loop interactions that may cause unknown biases which can be exacerbated after several iterations of machine learning predictions and user feedback. Such machine-caused biases risk leading to undesirable social effects such as polarization, unfairness, and filter bubbles. In this research, we aim to study the bias inherent in widely used recommendation strategies such as matrix factorization and its impact on the diversity of the recommendations. We also aim to develop probabilistic models of the bias that is borne from the interaction between the user and the recommender system and to develop debiasing strategies for these systems. We present a theoretical framework that can model the behavioral process of the user by considering item exposure before user interaction with the model. We also track diversity metrics to measure the bias that is generated in recommender systems, and thus study their effect throughout the iterations. Finally, we try to mitigate the recommendation system bias by engineering solutions for several state of the art recommender system models. Our results show that recommender systems are biased and depend on the prior exposure of the user. We also show that the studied bias iteratively decreases diversity in the output recommendations. Our debiasing method demonstrates the need for alternative recommendation strategies that take into account the exposure process in order to reduce bias. Our research findings show the importance of understanding the nature of and dealing with bias in machine learning models such as recommender systems that interact directly with humans, and are thus causing an increasing influence on human discovery and decision making.