DIMACS REU 2025 Participant
I am a student at Tsinghua University participating in the DIMACS REU 2025 program. My research interests include machine learning, causal inference, liquid democracy, social networks, and empirical economics. I am passionate about using computational methods to improve decision-making in social systems. This project on liquid democracy aligns with my interest in collective decision-making and governance.
Title: The Accuracy of Liquid Democracy
Mentor: Lirong Xia
Abstract: Liquid democracy is a hybrid voting system allowing voters to vote directly or delegate their voting power to others, who can further delegate. Its accuracy compared to traditional voting systems is underexplored. This project analyzes liquid democracy’s performance across network structures and competence distributions, using a stochastic model to capture delegation uncertainty. We aim to inform the design of robust liquid democracy systems.
I started my DIMACS REU project with mentor Lirong Xia, studying liquid democracy mechanisms. I explored basic concepts and set up a Python environment to simulate voting systems. I tested a simple network model and worked on calculating voter contributions. I faced issues with network setup, resolved through tool exploration. I planned a voter preference questionnaire. Mentor discussions helped shape my focus on mechanism comparisons.
I worked on implementing voting mechanisms and expanding my simulation. I coded systems to handle delegations and fixed issues with looping delegations. I developed a method to measure voter impact and refined a questionnaire for voter preferences. Challenges with setup were addressed by adjusting parameters. Mentor meetings guided my work on network effects. I had a working simulation by week’s end.
I prepared a symposium presentation.I ran simulations on different network types and improved my code for efficiency. I tested voting systems and measured voter contributions. I simulated voter preference data and prepared a progress report for a mentor meeting. Challenges with stability were fixed by tweaking code. I started gathering data for machine learning. The report and discussions shaped my next steps.
I explored machine learning to predict voter contributions, training a model with network features. I tested new directions, including changing networks and multiple voting issues. I faced model setup issues, resolved by adjusting settings. I drafted a final report. I will further exploring my project in the future.
Email: xuxz23@mails.tsinghua.edu.cn
Personal Profile: https://xuxz23.github.io/
I thank my mentor, Lirong Xia, for his guidance, and the DIMACS REU coordinators, Lazaros Gallos, Larry Frolov, and Martin Cerny, for their support. Thanks also to my fellow participants for their collaboration.