Facebook gave a large amount of anonymized data to a science project for research purposes that was not complete as expected. This could have an impact on an as yet unknown number of scientific papers on the influence of social media on democratic elections, reports US media.
The approximately 1 exabyte large data set should contain information on around 38 million URLs that Facebook users publicly shared at least a hundred times on the platform between January 2017 and July 2019. The project got the data from Facebook in 2020.
The scientific work of a good 100 researchers who are part of the Social Science One consortium founded in 2018 could be affected by the error. Facebook apologized for the incorrect data, reports the Washington Post.
Due to a technical error, Facebook is said to have excluded from the dataset of US members of its social network those who have no verifiable political leanings. That’s about half of the Facebook accounts in the US. Accounts from other countries should not be affected.
The Italian internet researcher Fabio Giglietto from the University of Urbino is said to be loud New York Times Discovered anomalies in the data set and brought them to the attention of Facebook. According to Harvard professor Gary King, who co-initiated Social Science One, dozens of papers are said to be affected by the problem. The extent and consequences of this could not be determined until Facebook provides a correct dataset, but that could take weeks.
The data set is said to be one of the largest that has ever existed in the history of the social sciences. Precautions should have been taken to ensure that the individual Facebook users cannot be identified from the posts.