The search engine manipulation effect (SEME) and its possible impact on the outcomes of elections
Citation: Robert Epstein, Ronald Robertson (2015) The search engine manipulation effect (SEME) and its possible impact on the outcomes of elections. Proceedings of the National Academy of Sciences of the United States of America (RSS)
Internet Archive Scholar (search for fulltext): The search engine manipulation effect (SEME) and its possible impact on the outcomes of elections
Tagged: Sociology (RSS)
Can the rankings of search engine results be manipulated to influence voter preferences for election candidates? Robert Epstein and Ronald E. Robertson sought to answer this question through a series of randomized controlled experiments. Their findings show that the preferences of undecided voters can be shifted towards a particular candidate by 20% or more.
Search engines are used by millions of voters during elections to learn about candidates. Previous work has shown that users tend to focus on websites that rank highly in search results, even if these websites are not necessarily related to what they search for. As such, if the rankings of websites that support a particular candidate are manipulated by a search engine to be higher in the results, then voters may receive biased information, potentially influencing election outcomes.
Epstein and Robertson conducted a series of randomized controlled experiments to test the effect of search engine manipulation on voter preferences. In the first three experiments, eligible voters from San Diego, CA were recruited to test the effect of such manipulation in a lab setting for a recent election in Australia. The third of these experiments was then replicated as a nation-wide, online experiment across the United States using Amazon's Mechanical Turk platform. Lastly, a large-scale field experiment was conducted in India during a real-world election.
In the first four experiments, voters from various demographics (Republican, Democrat, independent, etc.) read biographies for candidates from a recent election in Australia. After this, they spent 15 minutes searching for more information about these candidates using a mock search engine developed by the researchers. In each treatment group, the rankings of search results were manipulated to favor particular candidates. The final experiment followed a similar design, but it was conducted in India during a real election with real candidates. The demographics for this experiment reflected that of Indian politics.
In the first three experiments, there was a 48.4% increase among the treatment groups for subjects that would vote for the candidate with biased search rankings. Of these subjects, 75% of them did not detect that the manipulation was occurring. In the fourth experiment, there was a 36.7% among the treatment groups for subjects that would vote for the candidate with biased search rankings. Of these subjects, 91.4% of them did not detect that the manipulation was occurring. Lastly, in the field experiment in India, there was a 9.5% increase among the treatment groups for subjects that would vote for the candidate with biased search rankings. Of these subjects, 99.5% of them did not detect that the manipulation was occurring.
Bond, R. M., Fariss, C. J., Jones, J. J., Kramer, A. D., Marlow, C., Settle, J. E., & Fowler, J. H. (2012). A 61-million-person experiment in social influence and political mobilization. Nature, 489(7415), 295.
Fogg, B. J. (2002). Persuasive technology: using computers to change what we think and do. Ubiquity, 2002(December), 5.
Theoretical and Practical Relevance
Put together, these findings suggest that the manipulation of search rankings has a statistically significant effect on voters' likelihood to vote for a particular candidate. Although the magnitude of this effect varied among the five experiments, Epstein and Robertson postulate that it would be fairly easy for a search engine to nudge 20% of undecided voters towards a particular candidate. Thus, without transparency and regulation, search engines could be manipulated to effect the outcome of elections.