Research tools
Table of contents
Collaborative bibliography
The Mozilla Misinformation Initiative maintains an open database of research articles relating to information manipulation.
Social Science One
Social Science One(SS1) is an initiative founded by Facebook and the Social Science Research Council(SSRC), an American non-profit organization. Social Science One is based on the work of academics Gary King (Harvard Professor and Director of the Harvard Institute for Quantitative Social Science) and Nathaniel Persily (Stanford Law School Professor). It is incorporated as a USA non-profit LLC. It aims to “understand and ameliorate society’s greatest challenges” through social research based on useful private data, while maintaining end-user privacy.
It implements partnerships between academic researchers and private industry actors, granting the former access to relevant information amassed by the latter, who can “enlist the scientific community to improve their business and produce social good, while protecting their competitive positions”. As of August 2020, Facebook is the only publicized partnership and the main source of information on SS1.
The structure of the partnerships enables a commission of “trusted academics” to obtain “relevant company information and systems” and to invite other independent scholars to do “research in specific areas, following standard peer review protocols, funded by nonprofit foundations, and with no required pre-publication approval by the company”. In addition to accessing datasets, academics can apply for funding to study them.
These independent scholars must go through a peer review process before being allowed to access data and funding. Academics can apply for grants by answering formal Requests for Proposals (RFPs) issued by the Social Science Research Council (SSRC) and whose criteria are defined by SS1. The SSRC also leads the peer review process and manages the research along with SS1.
Twitter Health Metrics
Disinformation datasets
Several Twitter and Reddit datasets are searchable online on io-archive.org.
- 3 million Russian troll tweets. Interesting training can be found here.
- Elections Integrity official dataset: IRA, Iran.
- 200 000 Russian troll tweets by NBC.
- OSoMe datasets.
- https://github.com/eldinsahbaz/TwitterBotDetection/tree/master/DataSet
Classification student projects
- Best-documented training attempt of a neural network.
- Training attempt of a neural network with a Chrome plug-in.
- Classifier attempt.
Fake news
- Lead Stories detects the most trending stories from known fake news, satire and prank websites and tries to debunk them as fast as possible. Methodology.
