Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 112 (33), E4512-21

The Search Engine Manipulation Effect (SEME) and Its Possible Impact on the Outcomes of Elections

Affiliations

The Search Engine Manipulation Effect (SEME) and Its Possible Impact on the Outcomes of Elections

Robert Epstein et al. Proc Natl Acad Sci U S A.

Abstract

Internet search rankings have a significant impact on consumer choices, mainly because users trust and choose higher-ranked results more than lower-ranked results. Given the apparent power of search rankings, we asked whether they could be manipulated to alter the preferences of undecided voters in democratic elections. Here we report the results of five relevant double-blind, randomized controlled experiments, using a total of 4,556 undecided voters representing diverse demographic characteristics of the voting populations of the United States and India. The fifth experiment is especially notable in that it was conducted with eligible voters throughout India in the midst of India's 2014 Lok Sabha elections just before the final votes were cast. The results of these experiments demonstrate that (i) biased search rankings can shift the voting preferences of undecided voters by 20% or more, (ii) the shift can be much higher in some demographic groups, and (iii) search ranking bias can be masked so that people show no awareness of the manipulation. We call this type of influence, which might be applicable to a variety of attitudes and beliefs, the search engine manipulation effect. Given that many elections are won by small margins, our results suggest that a search engine company has the power to influence the results of a substantial number of elections with impunity. The impact of such manipulations would be especially large in countries dominated by a single search engine company.

Keywords: Internet influence; digital bandwagon effect; search engine manipulation effect; search rankings; voter manipulation.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Fig. 1.
Fig. 1.
Search rankings for the three experiments in study 1. (A) For subjects in group 1 of experiment 1, 30 search results that linked to 30 corresponding Web pages were ranked in a fixed order that favored candidate Julia Gillard, as follows: those favoring Gillard (from highest to lowest rated pages), then those favoring neither candidate, then those favoring Abbott (from lowest to highest rated pages). (B) For subjects in group 2 of experiment 1, the search results were displayed in precisely the opposite order so that they favored the opposing candidate, Tony Abbott. (C) For subjects in group 3 of experiment 1 (the control group), the ranking favored neither candidate. (D) For subjects in groups 1 and 2 of experiment 2, the rankings bias was masked slightly by swapping results that had originally appeared in positions 4 and 27. Thus, on the first page of search results, five of the six results—all but the one in the fourth position—favored one candidate. (E) For subjects in groups 1 and 2 of experiment 3, a more aggressive mask was used by swapping results that had originally appeared in positions 3 and 28.
Fig. 2.
Fig. 2.
Clicks on search results and time allocated to Web pages as a function of search result rank, aggregated across the three experiments in study 1. Subjects spent less time on Web pages corresponding to lower-ranked search results (blue curve) and were less likely to click on lower-ranked results (red curve). This pattern is found routinely in studies of Internet search engine use (–12).
Fig. 3.
Fig. 3.
Amount of time, aggregated across the three experiments in study 1, that subjects spent on each of the five search pages. Subjects spent most of their time on the first search page, a common finding in Internet search engine research (–12).
Fig. S1.
Fig. S1.
A possible synergistic relationship between the impact that search rankings have on voter preferences and the impact that voter preferences have on search rankings. The lower curves (red and green) show slow increases that might occur if each of the processes acted alone over the course of a year (365 iterations of the model). The upper curve (blue) shows the result of a possible synergy between these two processes using the same parameters that generated the two lower curves. The curves are generated by an iterative model using equations of the general form Vn+1 = Vn + r[Rn × (1 − Vn)] + r[On × (1 – Vn)], where V is voter preference for one candidate, R is the impact of voter preferences on search rankings, O is the impact (randomized with each iteration) of other influences on voter preferences, and r is a rate-of-change factor. Because a change in voter preference alters the proportion of votes available, its value in the model cannot exceed 1.0.

Similar articles

See all similar articles

Cited by 7 PubMed Central articles

See all "Cited by" articles

Publication types

LinkOut - more resources

Feedback