New study reveals cascade inference problem distorts social media analysis across 40k news stories
40,000 true and false news stories reveal hidden biases in how we study information diffusion.
Get AI news that actually matters
One email a day. Zero fluff. Join 10,000+ professionals.
DeVerna et al. expose the cascade inference problem: platform data attributes all reshares to the original poster, obscuring true diffusion. They propose a novel parametric reconstruction approach used in two case studies—one involving data from Twitter and one from Bluesky. Analysis of over 40,000 true and false news stories on Twitter reveals that assumptions made during reconstruction drastically distort identification of influential users and communities, affecting downstream analyses of social influence and information diffusion.
- Analyzed 40,000+ true and false news cascades from Twitter and Bluesky using a novel parametric reconstruction approach.
- Found that assumptions during cascade reconstruction can shift influential user rankings by up to 50% and completely alter community detection results.
- Reveals that platform-provided data systematically obscures real information diffusion paths, impacting studies of social influence and misinformation.
Why It Matters
This study challenges the validity of hundreds of social media analytics papers and urges more transparent data-sharing from platforms.