Exposure to Social Engagement Metrics

Synthesis produced by the Fondation Descartes of the following research paper:

Avram, M., Micallef, N., Patil, S., & Menczer, F. (2020). Exposure to Social Engagement Metrics Increases Vulnerability to Misinformation. The Harvard Kennedy School Misinformation Review.

Does the number of “likes” or “shares” received by an article that appears on our Facebook or Twitter news feed influence the trust that we place in it? Do these signs of positive engagement with an article lead us, in turn, to like it and share it with others?

These are the inquiries that the authors of this study sought to answer. To do so, they analyzed data from individuals playing Fakey, a game that simulates a Facebook or Twitter news feed. Players have the option to “share,” “like,” “fact-check,” or “skip” each article appearing on their news feed. Before participating in the game, players are told that sharing content means that they entirely support it, that liking it supports it, and that fact-checking warns that it may not be reliable. On the game’s news feed, half of the articles posted were from reliable sources, and the other half came from low-credibility media sources. ¹

The authors analyzed the behavior of approximately 8,600 players, to whom more than 120,000 articles were submitted over a 19-month period (May 2018 to November 2019). 78% of the players resided in the United States.

As expected, players are more likely to endorse content that has been widely liked or shared than content that has been otherwise ignored by other users. Third-party involvement with a post thus seems to function as an indication of credibility in the eyes of social network users, who will in turn be more inclined to share it themselves. Moreover, the authors note that the majority of misinformation campaigns exploit this phenomenon by using bots to like and share their content.

The authors conclude that online platforms should rethink the way in which user engagement surrounding content on these platforms is displayed, such that it no longer facilitates the spread of disinformation and does not disadvantage the diffusion of reliable information.

1. The authors note that this proportion of reliable to unreliable articles is not intended to realistically simulate what is observed on social networks, but that it allows for a better assessment of their users’ behavior.[⇡]