Abstract

The emergence of echo chambers across online social networks is boosting polarization of individual users regarding sensitive topics such as gun control, urban violence, racism, abortion, vaccinations, and police brutality. There is limited understanding of the influential nodes within such homogenous communities, and how echo chambers interact with each other. In this project, we will leverage tools from network science to shed light on the structures of echo chambers and detect their “leader” content creators. Results will help social media platforms in developing feed algorithms tailored towards dismantling chambers that are incubators for extremely polarized opinions affecting our society.

Project Description & Overview

Exposure to information from social media platforms has radically changed the way we discuss certain topics, including gun control, urban violence, racism, abortion, vaccinations, and police brutality. How users are connected in these platforms plays an important role on how information diffuses among users and on the type of information they are exposed to. Understanding the interplay between users’ connections and content shared in online platforms is the chief goal of this project.

Usually, we tend to accept and seek opinions that are similar to our beliefs. This tendency will likely be reflected in strong connections in online platforms between individuals sharing similar ideologies. Feed algorithms operated by platforms such as Twitter might also boost the creation of such homogeneous communities. Indeed, researchers showed the emergence in online social networks of “echo chambers” of users with similar opinions, which can boost polarization and extreme opinions.

In this project, we will study the structure of gun control echo chambers and the features of influential users (leaders). We hypothesize that echo chambers’ leaders are the ones that are mostly connected to other chambers, as they seek to feed the fire of debate.

We will collect Twitter data on the debate on gun control or other controversial topics. Using sentiment analysis and topic modeling, we will quantify users’ leaning toward the topics. Techniques from network science will be employed to 1) detect echo chambers and their leaders and 2) test our hypothesis. Statistical analyses will be performed to offer reliable and scalable conclusions.

Datasets

Tweets

• Tweets will be collected using the official Twitter API with the help of Python package Tweepy and through The Ohio State University software Hydrator.
• Older tweets will also be collected through web scraping using sns-scrap.
• Data from the Internet Archive Twitter Stream will be also utilized.

The Sentiment Lexicon dictionary from Pittsburgh University will be used for the sentiment analysis. 

The Natural Language Processing Toolkit on Python will also be utilized alongside the Vader sentiment dictionary.

Competencies

• Statistics
• Data extraction and web scraping (preferably Python)
• Data analysis and visualization (using Python, R or Matlab)
• Programming (preferably Python or Matlab)

Learning Outcomes & Deliverables

1. Students will learn data collection techniques and pre-processing methods.
2. Students will be trained on hypothesis testing and discover data modeling for analysis purposes.
3. Students will learn to apply common tools in network theory and community detection.

Results:

Interactions among Echo Chambers in Social Media

Students

Jiazan Shi, Siming Wu, and Yue Yang