Recent Awards

Information retrieval systems, such as search engines and recommender systems, are responsible for organizing and curating most of the information that is being consumed today. However, societal biases and stereotypes are reflected in search and recommendation results, and are often reinforced and amplified by information retrieval systems. Such biases can lead to under-representation of some social groups, but also to an unfair distribution of opportunities and resources. In this project, we will study and quantify various sources of bias in the state-of-the-art machine learning methods for information retrieval. Moreover, we will develop an algorithmic framework for formulating a broad range of fairness and accountability constraints on search and recommendation. We also aim to test and deploy the proposed approach in the industrial search and recommendation systems in order to improve fairness of these systems in the real world.