The Search Engine Manipulation Effect: Quantifying the Influence of Page Rankings
Relevant Articles: https://www.theguardian.com/commentisfree/2018/sep/06/google-search-results-rigged-news-donald-trump; http://www.pnas.org/content/112/33/E4512
At its core, the basic page rank algorithm we discussed in lecture is very simple. It’s an elegant solution to give pages scores based on how many in-links they have, find equilibrium scores among that network, and rank pages accordingly. Yet, it’s obvious that search engine optimization companies (SEO) like Google and Bing do much more than just use a standard page rank algorithm. I’ve begun to think critically about what motivates SEO companies to expand upon a simple page ranking algorithm. Of course, there are some great reasons such as trying to gain a competitive edge against other search engines or incorporating user data to get more personalized results.
However, there are other reasons. I’ve started thinking about the scale of the social influence that SEO companies possess. Search engines actively curate and filter information for us and have undeniable influence on what we ultimately see and don’t see via the internet. And with search engines being so commonplace now and at the fingertips of large percentages of entire nations, the influence search engines have on society can actually be massive. The article mentions studies that have shown that the manner in which page are ranked can have the same level of impact as stories reported by news/media outlets. The article brings up the interesting concept of search media, thinking of ranked search results as a form of media rather than just a service. There are even some studies referenced in the article that show that users will trust results that appear closer to the top of the page, even if the results originally at the bottom were swapped to be at the top. There is another study that refers to this as the search engine manipulation effect (SEMS) which quantified a search engine’s ability to shift public onion. One statistic that the study found was that search engines could shift the voting preferences of up to 20% of undecided voters. This is not insignificant. And what’s even scarier is that people are not aware that their opinions are being shifted by the search engine.
Thus, search engines have to be careful. A bare-bones page rank system has many issues as it could be susceptible to spam links or keywords from individuals or organizations seeking to game the system. However, the more a search engine adapts to problems like this, the more selective it becomes about information and the more of a say search engines have over what we should be seeing. And even if a search engine does this well and is impartial, it could still seem like it’s biased. I find this similar to the shuffle feature on music streaming devices. A truly random shuffle has a decent likelihood of grouping together related songs and thus can “feel less random.” To combat this, people have written algorithms to shuffle music in a manner that isn’t truly random but instead feels more random. For a search engine, a truly impartial search could still appear biased for many reasons. In the context of politics, perhaps a search engine prioritizes more reputable news sources and reputable news sources happen to be more left-wing. Then a page of impartial search results can appear to be incredibly politically biased to the left. This begs the question of whether search engines should change their search results to “appear” less-biased. But what if millenials tend to be more left wing and there are more millenials on the internet. Should search engines be providing a skewed representation of the number of opinions found online? This is a difficult question to answer.