VLDB 2024 Tutorial

Slides  |   References  |   PDF

Abstract

Given a large number (notationally m) of users’ (members or voters) preferences as inputs over a large number of items or candidates (notationally m), preference queries leverage different preference aggregation methods to aggregate individual preferences in a systematic manner and come up with a single output (either a complete order or top-k, ordered or unordered) that is most representative of the users’ preferences. The goal of this tutorial is to adapt different preference aggregation methods from social choice theories, summarize how existing research has handled fairness over these methods, identify their limitations, and outline new research directions.

PART I: Preference aggregation Method

In this part, we describe the basic model in computational social choice—voting. In voting, preferences are first elicited from an agent community, which are aggregated in the subsequent step. We formally define election, voting rule, input and output formats, preference aggregation, and analysis of the approach.

PART II: Fairness in answering preference queries - existing research

We model fairness by protected attributes. Each item/candidate is associated a set protected attributes. As an example, seniority level is a multi-valued protected attribute with three possible values Junior, Mid career, Senior, while gender is commonly a binary protected attribute with two values male and female. Building upon this, we summarize existing research about fairness in answering preference queries in several key aspects:

• ensuring fairness;
• multi protected attributes;
• producing a fair outcome.

PART III: Future research directions

We focus on three major aspects:

• new preference aggregation methods;
• alternative models to enable fair outcomes;
• efficient solution design.

Presenters

Senjuti Basu Roy is the Panasonic Chair in Sustainability and an Associate Professor in the Department of Computer Science at the New Jersey Institute of Technology. Her research focus lies at the intersection of data management, data exploration, and AI, especially enabling human-machine analytics in scale. Senjuti has published more than 85 research papers in high impact data management and data mining conferences and journals. She has served as the tutorial co-chair of VLDB 2023.

Baruch Schieber is a Professor in the Department of Computer Science, NJ Institute of Technology. Before joining NJIT, Baruch was a Distinguished Research Staff Member in IBM Research. His research interests are in theoretical computer science, including optimization under uncertainty, algorithms, mathematical programming, and high-performance computing. Baruch received his PhD in Computer Science from Tel Aviv University in 1987. He published more than 150 papers in scientific journals and conferences.

Nimrod Talmon is an Assistant Professor within the Industrial Engineering and Management Department, Ben-Gurion University, Israel. Before joining Ben-Gurion University, he was a Postdoctoral Fellow with the Computer Science and Applied Mathematics Department, Weizmann Institute of Science, Israel and received his Ph.D. degree in computer science from TU Berlin. His research interests include artificial intelligence, game theory, computational social choice, social networks, and combinatorial optimization. He has Erdos number 3, Sabbath number 7, and Bacon number 6.