An extensive study published in August 2015 in the Proceedings of the National Academy of Sciences USA shows that biased search rankings have a dramatic impact on the voting preferences of undecided voters. Five randomized, controlled experiments conducted with more than 4,500 participants in two countries showed that rankings that favored one candidate could easily increase the proportion of people who supported that candidate by 20 percent or more--up to 80 percent in some demographic groups--with virtually no one aware they were being manipulated. Because in most countries online search is conducted on a single search engine, this means that if, for any reason, search results on that search engine favored one candidate, a large number of votes would likely be driven to that candidate with no possible way of counteracting the effect. Because search algorithms currently do not incorporate "equal-time" rules to assure objectivity in presenting election-related material, and because many elections around the world are won by small margins, it is possible that a single search engine has recently been determining the outcome of upwards of 25 percent of the world's national elections, with increasing impact each year as internet penetration has been increasing.
The impact of search rankings on peoplei's thinking is called the Search Engine Manipulation Effect (SEME). SEME is one of the largest behavioral effects ever discovered, and it is almost entirely undetectable as a means of social influence, which makes it especially dangerous. Its impact extends far beyond voting, affecting decisions large and small that people make every day. SEME's power derives from a basic operant conditioning phenomenon: In routine searches every day, people are being trained, like rats in a Skinner box, to believe that what is higher in a list of search results is better and truer. The stronger that belief becomes, the more easily search rankings can be used to alter the beliefs, attitudes, opinions, and behavior of people who are undecided on almost any issue. In both scope and power, this makes SEME unlike any other list effect that has ever been discovered.
Ongoing research on SEME is assessing how the operant conditioning process contributes to SEME's power, how SEME is affecting decisions people make about their health, how SEME may be affecting court decisions, and how it might be possible to suppress SEME through regulations, browser add-ons, or other means.
The Stanford EE Computer Systems Colloquium (EE380) meets on Wednesdays 4:30-5:45 throughout the academic year. Talks are given before a live audience in Room B03 in the basement of the Gates Computer Science Building on the Stanford Campus. The live talks (and the videos hosted at Stanford and on YouTube) are open to the public.