On Tuesday, President Trump shared his opinion of Google:
CNN rapidly produced articles debunking the claim, including an editorial by Chris Cillizza in which he argues that this is a “conspiracy theory” and that Google’s search results have “very little to do with ideology.” Like other responses to President Trump’s tweet, Cillizza’s quotes Google, which in a statement reiterated its usual position: The search engine seeks to present material that is most useful to the user.
Much like Walter in The Big Lebowski, the president is not wrong in this case. Google is biased. It is designed to be.
A search engine’s main function is bias—that is, promoting some sources of information, while relegating others to obscurity. This, of course, is a process news organizations engage in every day: choosing sources according to a rubric that the reader does not fully know or understand. The fact that I do not know how a journalist has chosen her sources does not, in itself, make her story ideologically inflected, and it does not mean that it is either invalid or useless. I trust her professional training, and the checks of the organization she works for, to provide me with a useful picture of the news. And I can balance the potential bias of a story or a tweet against other information I find online. Likewise, I am not a climate scientist, but I trust the individual scientists and the structures they work within to provide me with useful answers, and can balance their conclusions with other peer-reviewed literature. At the core of these processes are two things: usefulness and trust.
What makes an information source useful is very much dependent on context. Tuesday morning, my wall clock stopped working, and I Googled “fix wall clock.” Google helpfully provided videos and websites that will guide me step-by-step in repairing the clock. And it is likely that the same search results would be useful to anyone who used those search terms. When it comes to more complex issues, and complicated motivations for searching, things get difficult.
We wouldn’t expect, for example, a search engine to effectively guess which upcoming movie we most want to read about. Someone who is excited by Keanu Reeves engaging in a horse-top shootout in John Wick 3 probably doesn’t care that the lead actor is being swapped in Greta Gerwig’s upcoming version of Little Women. (I eagerly await President Trump’s critique of the Tomatometer.) If we cannot accurately predict movie preferences, how do we expect the same search results to be useful to different people seeking information about immigration or monetary policy?
The response to this is likely that one is related to preferences, and the other is related to facts. The oft-heard critique of news bias is that journalistic outlets should simply report the facts. But there is nothing simple about choosing which facts to include, arranging them in a narrative, and then selecting which should be presented first. This is an art that has been perfected over decades by many editors. How could the Google algorithm have already perfected it?
In fact, we know that it hasn’t. Were the Google ranking algorithm perfected, it would not continue to change over time, and as best we can tell, it changes fairly frequently in large and small ways. It is altered not just because it has, on its own, somehow detected that its search results are not the most useful, but sometimes thanks to external criticism. A decade ago, when a white supremacist–produced site became one of the top hits for people looking for information about Martin Luther King Jr., it was seen as a flaw in the algorithm; thanks to changes in the way it weighted factors of credibility, the site was displaced from the top ranks (though it continued to appear in deeper results for some time). The site it presented was not entirely factual, but more importantly, it presented a position that was acceptable to only a small minority of Google users. Thanks to a change in the ranking algorithm—an algorithm written by humans—that site no longer shows up among the highest-ranked results.
The question then becomes how much we should trust an algorithm that has already been adjusted under public pressure. There is an ideological bent to the search algorithm, and it is the suggestion that Google is somehow a neutral, natural, mathematical process by which we may reach the “best” results for our searches. There is no natural or neutral selection of news sources that will make all users happy. Being suspicious of claims that a particular order of importance is somehow natural—especially an importance that is derived from Delphic black box like the Google ranking algorithm—is not a conspiracy theory, it is sensible skepticism.
There is a public interest in a democracy providing its citizens access to information. It is the reason we have the Library of Congress and the reason the government still subsidizes the mailing of printed matter. Even if Trump’s claim isn’t exactly wrong the tweets are still jarring, particularly because of the final words: “This is a very serious situation-will be addressed!” The constitutional protection of free speech in the United States has meant that the idea of regulating sources of information has traditionally been a no-go zone. If the president does not like the information found on Google, on Facebook, or on Wikipedia, he (though he barely uses computers) and his like-minded constituents are free to vote with their mouse clicks and choose alternative sites where alternative facts are given higher billing.
The problem, of course, is that Google is by far the most popular search engine in the United States, and the infrastructure required to create good search represents a substantial challenge for rivals—it is, in some ways, a natural monopoly. Other countries have seen a U.S.-based global search engine as a challenge to their countries’ cultures and governments, and have funded attempts at alternatives. The Chinese have been particularly effective in this regard, as Baidu garners a rapidly growing user base both within the country and abroad. But no one comes close to Google, which provides results for, by one estimate, more than 90 percent of global searches.
We do not have a similarly monopolistic news outlet. Those who do not like the coverage of the Washington Post, or the “mainstream media” writ large, have alternatives that the president has actively endorsed, from Fox News to Infowars. Indeed, access to alternatives is easier now than it has ever been. But the choices of search engine are much more restricted.
Monoculture when it comes to filtering content is dangerous, not least because it makes the potential for intentional and unintentional abuse high.
As Supreme Court Justice Louis Brandeis suggested nearly a century ago, “to avert the evil by the processes of education, the remedy to be applied is more speech, not enforced silence.” President Trump has a powerful tool for driving the search engine market: He could push for public research funding. The dominance of Google makes its search algorithms assailable mainly by efforts funded by other governments, and well-heeled rivals like Microsoft, Facebook, and Amazon. Were the National Science Foundation to commit substantial new funding toward public research on improving search algorithms, no doubt the president would find among the thousand blooming flowers an alternative that would better match his view of what is important.