CIB Mango Tree
A Civic Tech DC Project
About Download & Install Data Preparation Test Library
Post Time Coordination Test This test scans for usernames that post at similar times on more than one occasion, revealing signs of possible coordination. It will rank groups of usernames by the number of times that their posts appeared together during the timespan of your dataset, from highest to lowest. You will be able to identify whether sets of usernames are posting together at a frequency that suggests coordination or bot activity, and examine their content. See Code for Test on GitHub Magelinski, Thomas and Kathleen Carley. "Detecting Coordinated Behavior in the Twitter Campaign to Reopen America." in Social-Cybersecurity in Times of Crisis and Change Conference, 2020.
Unique Post Username
Unique Post Number
Date
Post Time
Post Content

  1. The test will divide the timespan of your dataset (the length of time over which data was collected) into discrete 10 minute windows.
  2. The test will tally how many times a pair of usernames posted within the same 10-minute window on multiple occasions throughout the duration of your dataset.
  3. In order to check for larger groups that may be coordinating, the test will then count the number of times a trio of usernames posted in the same 10 minute time window in more than one instance, followed by the number of times a quartet (4) usernames posted in the same 10 minute time window, and so on.
  4. The test will cease running when the maximum group size to have posted in the same 10 minute time window in more than one instance is found.
  5. The test will then rank these groups by the number of times they posted in more than one time window, from highest to lowest. The test will produce a separate csv file for each group size (pairs, trios, etc.) that was found.

.csv Output Format
  • The test will produce a separate .csv file for each group size that was found to have posted in more than one time window. (NOTE: The test will only produce a csv file for group sizes found. I.e. there may not necessarily be groups of size three or four which posted in the same time window over more than one occasion. Therefore, the test may not produce a csv displaying groups of size three, four etc.)
  • Within each csv file, the test will rank each group by the number of times they posted across more than one time window, from highest to lowest. The csv will display all the instances that the usernames belonging to each group posted, including the UPNs of each post, the content of each post, and the time of each post
csv file