Tracking the Online Spread of Misinformation after Disaster Events
Directed Research Group, UW HCDE
Jan. 2015 - Jun. 2015
Research Goal:
-
Analyze quantitative/qualitative data on tweets during disaster events to track misinformation
-
Study and understand the collective sense making process in the propagation of misinformation
My Contribution:
-
Disaster Event Identification
-
Rumor Identification and Scope Definition
-
Data Analysis - Qualitative Data Analysis
-
Data Analysis - Quantitative Data Visualization and Analysis
Iterative Process
-
Disaster Event Collection
-
Rumor Identification
-
Tweet Categorization
-
Data Analysis
-
Automated Rumor Detection
Publication:
-
Starbird, K., Spiro, E., Arif, A., Chou, F., Narisimhan, S., Maddock, J., Shanahan, K. & Robinson, J.(2015). Expressed Uncertainty and Denials as Signals of Online Rumoring. Collective Intelligence.
-
Arif, A., Shanahan, K., Chou, F. J., Dosouto, Y., Starbird, K., & Spiro, E. S. How Information Snowballs: Exploring the Role of Exposure in Online Rumor Propagation. ACM 2016 Conference on Computer Supported Cooperative Work
1. Disaster Event Collection
-
Keep track of latest news or trending events on Social Media
-
Identify disaster event and the keywords related to the event
-
Collect real-time tweets during the event and Store them in to Mongo DB
2. Rumor Identification
-
Identify rumors during an event
- using visual pattern analysis with examination of external sources -
Develop the definition, scope and search string for the rumors
- using an iterative process to refine search strings
3. Tweet Categorization
-
Code the tweets in a certain rumor as affirming the rumor, denying the rumor, or showing uncertainty about the rumor.
-
Three trained coders code all the tweets and use "Majority Rule" for adjudication. (i.e. if more than two out of three coders coded a tweet as "affirming the rumor", then we categorized this tweet as "affirm".)
4. Data Analysis
-
Analyze the quantitative and qualitative attributes of the tweets and look for trends and insights in collective sense making process on social media
-
Visualize information to help identify obscure trends
5. Automated Rumor Detection
-
Use the information from data analysis to inform automated methods of detecting rumors on social media platforms during crisis events.
