Detect Coordinated Inauthentic Behavior on a National News Platform
Overview
What this challenge is about.
Receive an anonymized 90-day export of comments (user_id, article_id, timestamp, like_count, reply_to_id) and basic user metadata (registration date, login pattern bucket). Build a user-user co-engagement graph (edge = co-commented on N articles within K minutes), apply community detection (Louvain or Leiden), and compute coordination features: temporal synchrony, semantic overlap (sentence embeddings via a small open model run locally), and structural roles (hubs, brokers). Score each user 0-100 for coordination likelihood and prototype a moderator dashboard. Deliver: 16-page analysis, scoring layer prototype (Python notebook + small Streamlit UI), 8-page moderation playbook.
The Brief
What you'll do, and what you'll demonstrate.
Identify coordinated inauthentic users in a 1.8M-comment dataset using network science and ship a scoring layer the trust-and-safety team will act on.
Earning criteria — what you'll demonstrate
- Construct a co-engagement graph from raw event data
- Apply community detection on a 100k+ node graph
- Combine structural and semantic features for behavior scoring
- Translate a network model into operational moderation policy
Program Fit
Where this fits in your program.
Sharpens the same skills your degree expects you to demonstrate.
Skills
Skills you'll demonstrate.
Each one shows up on your verified credential.
Careers
Roles this prepares you for.
Real titles. Real skill bridges. Pick the one closest to your trajectory.
Product Manager
Product managers who can read a network model and write the moderation policy that goes with it become the trusted PM for integrity teams.
This challenge sharpens
- computational-social-science
- trust-and-safety
- data-analysis