2021-2022 PRojects
Project Title: Theoretical Models for Opinion Polarization in Social Networks
Professor: Cameron Musco
Lab/Research Group: Theory Group
Members of the Theory Group use mathematical techniques to study problems throughout computer science. Our work is extremely diverse -- it includes graph and network algorithms, randomized approximation algorithms, streaming algorithms, combinatorial optimization, computational geometry, dynamic algorithms and complexity, model checking and static analysis, database theory, descriptive complexity, parallel algorithms, and computational complexity theory. Members of the Theory Group wear other hats as well and collaborate throughout the department and the world beyond. Every week we all get together for a research seminar, featuring either one of our own professors or Ph.D. students, or a visitor from another institution. You can find more about the group and our weekly seminars at https://groups.cs.umass.edu/theory/.
Recently, significant attention has focused on increased polarization of political opinions. It is thought that this polarization may be tied to algorithmic content recommendations on social media platforms, which can lead to so-called 'filter bubbles', where extreme opinions are continuously validated and not challenged by diverse viewpoints.
This project hopes to explore these issues by considering opinion dynamics and polarization on simple random graphs, representative of real work social networks. Using standard opinion dynamics models from the social science literature, like the Friedkin-Johnsen model, we will explore how graph structure, various forms of algorithmic recommendation, and other factors, affect opinion polarization. We will try to find basic modifications of existing recommendation methods that limit the filter bubble effect and associated polarization.
The majority of the work will be through simulations on synthetic random graphs and real world social networks seeded with synthetic opinions. Students will have the opportunity to conduct background research on and build implementations of existing opinion models and basic content and link recommendation systems. The work will build students' comfort with numerical simulation and network analysis in Python or Matlab. There may also be room for some theoretical investigation for interested students.
