The principal research objective of the project is to evaluate deep learning methodologies using neural networks for predicting information diffusion processes in various social online environments. While deep learning has been shown to be a valuable tool in recognizing images, it has not been sufficiently explored in the context of dynamic processes on social networks. Yet, we believe, with the right techniques in place, deep learning can contribute significantly to predicting dynamic processes on social networks at scale.


Neural networks are a computational model based on a large collection of simple neural units (artificial neurons) connected with many others via weighted links. Each individual neural unit aggregates its weighted inputs and the transfer function can activate or inhibit the state of linked neural units. The activation/inhibition is controlled by the weights on links. Neural networks are self-learning and trained, rather than explicitly programmed, and excel in areas where the solution is non-linear or feature detection is difficult to express in a traditional computer program.  Deep neural networks typically consist of multiple layers of connected neural units that are neither input nor output, called hidden layers. Deep learning has proven effective in domains where it is possible to collect lots of data.


The novelty of our approach consists in naturally addressing concerns related to the tradeoff between complexity and scale in the simulation of information diffusion in social networks. What this project will provide is a rigorous investigation of what input is required for what level of simulation accuracy. In this process, we will specifically address the following challenges:

  • Representing and linking properties and behaviors of populations at different granularity (e.g., from aggregate dynamics of large communities to discrete interactions at small scale).
  • Allowing for inherently complex behavior at the individual or group level while allowing for a scalable simulation to millions individuals.
  • Simulating not only the spread of immutable information, but also the spread of information as it evolves in its diffusion
  • Simulating information spread and evolution across multiple environments.
Publications
  • Online Discussion Threads as Conversation Pools: Predicting the Growth of Discussion Threads on Reddit
  •      Computational and Mathematical Organization Theory, 2021.  [paper]
  • Simulating Temporal User Activity on Social Networks with Sequence to Sequence Neural Models
  •      IEEE International Conference on System, Man and Cybernetics, 2020.   [paper]
  • Using Deep Learning for Temporal Forecasting of User Activity on Social Media: Challenges and Limitations
  •      WWW'20: Companion Proceedings of the Web Conference, 2020.   [paper]
  • Mentions of Security Vulnerabilities on Reddit, Twitter and GitHub
  •      IEEE/WIC/ACM International Conference on Web Intelligence, 2019.   [paper]
  • Predicting Longitudinal User Activity at Fine Time Granularity in Online Collaborative Platforms
  •      IEEE International Conference on Systems, Man and Cybernetics (SMC), 2019.   [paper]
Pre-prints (under review)
  • Simulating Fine-Grained Twitter Activity during Political Crisis Using Exogenous Data
  •      (manuscript under revision).   [repo]

This project is an interdisciplinary collaboration between Computer Science and Sociology. Our team includes three faculty, one postdoctoral researcher, 5 graduate students and 2 undergraduate students.

Team
  • Frederik Mubang

    PhD Student

  • Kin Wai NG

    PhD Student