Monitoring Trending Disinformation Content on Facebook

The next speaker in this ICA 2024 conference session is the excellent Giada Marino, presenting some of the work of the Vera.ai research project. Responding to the challenge of mis- and disinformation, the project focusses especially on the coordinated communication networks that share such content in order to influence and manipulate social media audiences, and has developed a content-agnostic tool that monitors the activities of identified problematic accounts.

This currently works with CrowdTangle data, and is therefore under threat from the impending closure of the CrowdTangle data. Monitoring lists of problematic accounts derived from Meta alerts, it gathers up to 100 overperforming posts from these accounts every six hours, and identifies the most widely shared posts, most widely commented posts, and any coordinated links in these datasets.

AI is used to generate labels for the list of accounts that shared the coordinated links, and the initial list of accounts is updated with any newly discovered accounts that are engaged in coordinated link-sharing activities. The results of this analysis are then stored in a Google Sheet and posted to a private Slack channel.

The identification of these posts by the alert system then makes it possible to generate maps of frequent coordinated account activities. Typical at present here are online gambling spam networks, as well as political networks from South America, Africa, the US, and Europe, of varying sizes. One particularly odd approach here is the frequent sharing of supposed ‘band/musician’ pages with gory or creepy profile images on Facebook through the ‘currently listening to’ function, without any further comment.