Identification and Examination Of Facebook Filter Bubbles Via Big Data Approach

Awarded the 1st place for best overall quality among 30+ competing projects in the NTU CS+X Hackathon

Research Question

  • On social media platforms, do people discuss politics mostly with those within the same filter bubble, i.e., who share the same political stances?

3D Visualization of Facebook Filter Bubbles

Scroll your mouse wheel (or slide two fingers up or down on your trackpad) to zoom in and out.
Drag to points to see each person's cluster label. Drag while clicking to rotate the plot.

    Brief Methods

  1. We obtained Taiwanese users' political liking data on political posts, an indicator for political attitude towards major controversial social issues, by web-crawling in Javascipt.
  2. Ultimately, more than 19 millions of users' post-liking data were crawled (with 342,796 unique users).
  3. We performed k-means clustering to cluster users in R program. Nine filter bubbles were found, as shown above. 3D Interactive visualization was built by the plotly library in the R environment.

    Main Results

  1. The boundary of Facebook users filter bubble is formed by users' stances on their support for major political parties. specific social issues , consistent with young voters' political attitude.
  2. Generally, users within each filter bubble were highly isolated to one another, i.e., they predominantly only care about the issues they concern.
  3. One noteworthy exception is that the the three following filter bubbles overlap, i.e., Pro-Same-Sex Marriage, Anti-Death-Penalty, and Pro-Independence. That is, these social attitudes usually tie together.

    4. The links to the presentation slides and github repository