Google could be secretly influencing the minds of voters around the world to rig elections without anybody realizing it. No, that is not just a conspiracy theory, it is a serious argument put forth by Robert Epstein of the American Institute for Behavioral Research and Technology, a research organization based in California. In a recent study, Epstein and his team found that in a highly controlled setting they could manipulate the percentage of study participants that favored a particular candidate by at least 20 percent by changing the order of webpages listed in search results, without participants catching on—a phenomenon they call the “search engine manipulation effect.” Epstein concludes that based on the importance of undecided voters in elections, Google could exploit this effect to create specific election results—a claim trumpeted widely by the media in recent weeks (see 1, 2, 3, 4). However, a closer look at the facts shows that there is little reason to worry that search engine bias poses a realistic threat to democracy warranting government intervention, especially considering just how little a role search engines actually play in shaping public opinion.
In an article Epstein penned in Politico to publicize his research, he argues that Google could deliberately tweak its search algorithms to order search results in a way that portrays a particular candidate more favorably and thus manipulates undecided voters into supporting that candidate. Any manipulation of voter preferences in a laboratory setting typically only affects subjects for a short period of time, but over time, asserts Epstein, the impact of this manipulation could grow as voters encounter biased rankings multiple times. Epstein alleges that given Google’s ties to prominent Democrats, either its executives or perhaps just a rogue employee could swing the election in Hillary Clinton’s favor.
Naturally, Epstein acknowledges that he has no reason to suspect any wrongdoing. As Amit Singhal of Google’s search team points out, Google has never re-ranked search results to influence user sentiment, political or otherwise. In fact, Google previously resisted high-profile calls to alter its search results when past (and current) presidential candidate Rick Santorum asked the company to address a campaign led by blogger Dan Savage to associate a sexual neologism with his name. Moreover, manipulating search results for political gain could seriously undermine user confidence in the company. It is hard to imagine any company sabotaging its long-term financial prospects for short-term political gain.
However, a lack of evidence does not make Epstein’s claim impossible–a hallmark of a good conspiracy theory. But it is highly improbable. Notably, Epstein fails to account for the fact that voters do not make political decisions in a vacuum. Given how heavily political information saturates newspapers, television, and other media during an election season, it is highly unlikely that search results would be the deciding factor in a voter’s mind before he or she goes to vote.
In fact, search engines may not have a substantial impact on political opinions at all because, according to a recent study, they likely only play a small role in providing political information. According to the study, 88 percent of U.S. adults get news directly from news organizations, such as by watching a news channel or reading a newspaper, 65 percent get news via word of mouth, and just 51 percent reported getting their news from search engines. For political news specifically, no participants reported to turning to search engines for information, with the majority preferring 24-hour cable news networks. While some individuals could be exposed to biased search results, it is much more likely that these individuals would be more influenced by other media. Moreover, news is just one of many factors that could influence someone’s political opinion—age, family life, socioeconomic factors, and peers all contribute to shaping a person’s political beliefs.
Despite the minimal role search engines could play in shaping politics, Epstein ultimately calls for government regulation of Google’s search algorithms to preserve the integrity of elections. But what could this kind of oversight possibly entail? Should search companies have to wait for regulators to approve every alteration to their algorithms? Moreover, search results are the product not only of algorithms, but also online content and user behavior. Would the government try to regulate those as well? Cable news programs have a larger influence on voters and can have much more obvious political biases, but there are few calls to regulate them because doing so would be a violation of free speech.
Government regulation of search engines could actually inhibit a fair democratic process, as it would necessarily slow the rate at which search companies could improve their algorithms that connect voters with information. Search engines expose voters, undecided or otherwise, to millions of pieces of information that could help them make up their mind about a political candidate. Prior to the existence of search engines, voters had to make decisions about political candidates based on whatever sources happened to be accessible, such as television or radio stations, newspapers, or peers, and with substantially less ability to fact-check or seek out differing opinions. But today, as search algorithms get better at answering users’ questions, voters can investigate issues with less effort. Governments should leave search companies free to develop these algorithms, which make for a more informed electorate, rather than hamstring them out of unsubstantiated fear that they might be undermining democracy.