The “digital personalization effect” (DPE): A quantification of the possible extent to which personalizing content can increase the impact of online manipulations

In recent published reports on the “search engine manipulation effect,” the “targeted messaging effect,” and the “answer bot effect,” exposure to biased content produced significant shifts in the opinions and voting preferences of undecided voters. In the present study, these effects were replicated on simulations of the Google search engine, X (f.k.a., Twitter), and Alexa, and biased content was also personalized. Participants (all from the US) were first asked to rank order news and other sources according to how much they preferred them. Then they were randomly assigned either to a group in which content about the 2019 Australian election for Prime Minister would be received from highly preferred or least preferred sources. Participants were also randomly assigned either to a group in which the content was highly biased to favor Candidate-A or to favor Candidate-B. In all three experiments, the voting preferences of participants who saw biased content – that is, content that favored Candidate A or Candidate B – apparently coming from least preferred sources shifted by relatively small amounts (17.1% in Experiment 1, 21.8% in Experiment 2, and 39.3% in Experiment 3); whereas, the voting preferences of participants who saw that same biased content apparently coming from highly preferred sources shifted by significantly and substantially larger amounts (67.7% in Experiment 1, 71.9% in Experiment 2, and 65.9% in Experiment 3). All shifts occurred in the direction of the candidate favored by the bias. We conclude that personalization of biased content can greatly increase the impact of such content.