| coverY | 0 |
|---|
The geoLab/CCI twitter dis/misinformation competition is part of an experiential learning project designed to give participants real-world experience in data security. We're focused on exploring how treating dis/misinformation as data poisoning could potentially help us identify intentionally misleading tweets this year.
This year we'll be hosting a three-phase contest. In phase 1, you'll be provided with a dataset of known tweets tagged as either being from credible sources, or sources known to spread dis/misinformation. Students will work to build the best deep learning model they can to disambiguate these cases, and will have access to prebuilt examples, a full tutorial, a live website they can submit their results to, and regular weekly office hours.
In the next two phases, we will be designing a second model on the basis of manually-identified mis/disinformation examples from Russian sources. Led by Dr. Anthony Stefanidis, this will be an opportunity to work on a research paper that explores the capability of machine learning to identify such cases. You will be working closely with us to both collect data (Phase 2) and produce models (Phase 3). At the end of April, we'll hold a second round of competition based on the labeled dataset you produce.
Feb 2nd, 2023: Kickoff
Phase 1: Feb 2nd - March 5, 2023: Disambiguition of sources on Twitter based on content of tweets
Phase 2: March 5 - March 31, 2023: Labeling of dis/misinformation
Phase 3: April 1 - April 30, 2023: Modeling of dis/misinformation labeled dataset
First week of May: Final Presentations