Background
Social media platforms have become an integral part of our daily lives, offering users a powerful means to connect and share information. Understanding how information propagates through social networks, influenced by user interactions and online recommendation algorithms, is crucial. For instance, such understanding can help companies devise effective marketing strategies or enable policymakers to develop mitigation measures against the spread of misinformation.
A common approach to studying information spread is through mathematical models that simulate propagation on social networks. Prominent examples include the Independent Cascade, Linear Threshold, and variations of the well-known SIR model. These models often rely on a graph structure where each node represents a user, and edges denote friendships or follow relationships. Each node’s state, such as being exposed to a piece of information or not, evolves based on predefined (often stochastic) rules that simulate user interactions.
Research Gap
These propagation models are typically developed inspired by a mix of social science experiments, hypotheses, and intuitive reasoning. While their simplicity has made them popular, little work has been done to rigorously assess how well these models reflect real-world information propagation. This project seeks to address this critical gap by exploring the fundamental question: How accurate are information spreading models?
Aim and Tasks
This project will evaluate the accuracy of existing information spreading models by comparing them to real-world data on information propagation.
Task 1: Real-World Data to Propagation Patterns
Analyze existing real-world data, such as tweets, retweets, likes, comments, and user connections (from datasets like https://github.com/KaiDMML/FakeNewsNet), to extract propagation patterns within networks.
Task 2: Testing Model Accuracy
Using the propagation patterns from Task 1, evaluate the accuracy of various existing information propagation models. This will involve collecting and implementing relevant models and selecting appropriate hyperparameters.
Task 3: Developing a New Model
Based on the findings from Task 2, propose new models that more closely align with real-world propagation patterns.
Requirements
The ideal candidate should have a solid understanding of graph theory and algorithm design and feel comfortable working with graph-based and textual data, using network and NLP techniques. While you may use any programming language, Python is strongly recommended.
Project’s Scope
This project is suitable for a 12-unit student project or a 24-unit honors project over a full academic year.
Contact
Supervisors: Ahad N. Zehmakan and Jing Jiang
If you are interested, please email ahadn.zehmakan@anu.edu.au with:
- What aspects of this project interests you the most.
- The type of research project you are seeking (6-unit, 12-unit, or 24-unit)
- A copy of your transcripts and/or CV
- Any question you may have