The Video Manipulation Effect (VME): A quantification of the possible impact that the ordering of YouTube videos might have on opinions and voting preferences.

Recent research has identified a number of powerful new forms of influence that the internet and related technologies have made possible. Randomized, controlled experiments have shown, for example, that when results generated by search engines are presented to undecided voters, if those search results favor one political candidate over another, the opinions and voting preferences of those voters can shift dramatically-by up to 80% in some demographic groups. The present study employed a YouTube simulator to identify and quantify another powerful form of influence that the internet has made possible, which we have labeled the Video Manipulation Effect (VME). In two randomized, controlled, counterbalanced, double-blind experiments with a total of 1,463 politically-diverse, eligible US voters, we show that when a sequence of videos displayed by the simulator is biased to favor one political candidate, and especially when the "up-next" video suggested by the simulator favors that candidate, both the opinions and voting preferences of undecided voters shift dramatically toward that candidate. Voting preferences shifted by between 51.5% and 65.6% overall, and by more than 75% in some demographic groups. We also tested a method for masking the bias in video sequences so that awareness of bias was greatly reduced. In 2018, a YouTube official revealed that 70% of the time people spend watching videos on the site, they are watching content that has been suggested by the company's recommender algorithms. If the findings in the present study largely apply to YouTube, this popular video platform might have unprecedented power to impact thinking and behavior worldwide.