PageRank as a Source for Bias in Search Engines
Tampering with public access to information and corporate or government censorship has been a concern for philosophers and politicians for hundreds of years. With recent news revelations in terms of Google’s search engine filter, the topic has been raised again, but with a different twist: if news is “fake”, is censorship of this kind still detrimental to society?
The Google search algorithm is designed to filter search results based on numbers of visits among other factors. In recent events the filter has been modified to account for sources that are presenting fake information in the form of news and other programs. The system runs like the PageRank system discussed in lecture: each site has a rank that goes up based on how many nodes are connected to it, and what each of their respective “rank” is. So, if a new site hosting erroneous data is both popular and often linked-to, the ranking would increase. The basic idea of the recent renovation is to lower the rank of devious news websites. However, an alteration in the PageRank system can effect other, “harmless” websites and a large alteration to the system is likely going to produce a fair number of false positives. Moreover, the vetting of information can demote content in specific groups that deviate from “official” or “reliable” sources. An example of this is the negative effect these changes have had on the world socialist guild, an institution that does not promote fake news, but is also not closely aligned with the mainstream news sites that are promoted in the new search algorithm. This algorithm is an example of how PageRank can be used to alter the user’s available content.
These results seem unsettlingly close to censorship mechanisms that could very easily spiral in any of a number of bad directions. It is arguable that the current vetting of information through any kind of search engine is, to an extent, this same sort of censorship, but (as far as has been disclosed) those do not seek a “truth” and hence an inherent “value” in a website, and thus no inherent or potentially political biases exist. If a site is strictly avoided because of its content, however, it is very questionable that such a filter can maintain the unbiased nature search engines strive to fulfill.
Source: https://www.nytimes.com/2017/09/26/technology/google-search-bias-claims.html