Multiple searches increase the impact of similarly biased search results: An example of the “multiple exposure effect” (MEE)

Abstract

A series of experiments published in the Proceedings of the National Academy of Sciences in 2015 showed that search engine results favoring one candidate can (a) shift the preferences of undecided voters by up to 80% in some demographic groups and (b) be masked so people show no awareness of the manipulation. We labeled this phenomenon the Search Engine Manipulation Effect (SEME), and it appears to be one of the largest behavioral effects ever discovered. In two follow-up experiments with a total of 1,032 undecided, eligible U.S. voters (mean age = 32.6), we have now replicated SEME with a different election (the 2016 Clinton/Trump election) and different conditions. All previous SEME experiments allowed subjects just one search opportunity. In the new experiments, we asked whether multiple searches on the same topic (but using somewhat different search terms) would shift voting preferences more than a single search would. In both of the new experiments all subjects were first shown brief biographies of the candidates, then asked about their voting preferences, then allowed to conduct an online search using our mock search engine (Kadoodle), and then asked again about their voting preferences. In both experiments, subjects in different groups saw search rankings favoring Mr. Trump, Mrs. Clinton, or neither. In Experiment 1, all search sessions lasted a maximum of 15 minutes, as in previous experiments. In Experiment 2, all search sessions were limited to 5 minutes, which is more typical of real search behavior. Our primary dependent variable was Vote Manipulation Power (VMP), the percentage increase in the number of subjects inclined to vote for one candidate after having viewed search rankings favoring that candidate. In Experiment 1, a VMP of 11.5% was found for groups conducting just one search. In the multiple-search conditions, the VMP increased with successive searches from 14.3% to 20.2% to 22.6%. In Experiment 2, a VMP of 8.0% was found for groups conducting just one search. In the multiple search conditions, the VMP increased with successive searches from 9.6% to 19.3% to 25.3%. Corresponding shifts were also found for how much subjects reporting liking and trusting the candidates and for subjects’ overall impression of the candidates. Because multiple, short searches are typical of user behavior, we conclude that our previous reports about the possible impact that similarly biased search results might have underestimated the impact that biased search rankings can have on people conducting multiple searches over time. Findings in the multiple-search experiments exemply what we call the “multiple exposure effect” (MEE); we summarize data from other studies which also appear to demonstrate the cumulative effects of multiple exposures to similarly biased online content.

Keywords: search engines, Search Engine Manipulation Effect, SEME, online manipulation, multiple search, Vote Manipulation Power, MEE, Multiple Exposure Effect