Google Search Algorithm
https://www.cnbc.com/2018/09/17/google-tests-changes-to-its-search-algorithm-how-search-works.html
Most of the general public do not quite understand the workings of Google’s search algorithms and for good reason. Logically, people will scrutinize Google for having bias for their own products. This lead to a $2.7 billion antitrust fund for its shopping results. They have also been under attack from President Trump for “showing political bias” in their search results. To debunk the accusations of bias, Google invited CNBC, a news channel, to sit in at a meeting. The executives were deciding whether or not to approve images near their search results. When someone types their input into the search bar, the search engine looks for pages where the words and their synonyms appear, surfacing relevant pages. “Web crawlers” are constantly going through billions of webpages storing that data and taking note of the freshness of the sites and where its located. When testing new algorithms for searches, Google takes a small number of real users and a group of contractors called “search quality raters” to see how they interact with it. They then asses the quality of the search according to a set of guidelines which value among other things, expertise and trustworthiness. In one of their latest tests, they wondered if displaying pictures alongside a link would help users better discern if the results matched their query. After the experiment, 91% of the time raters found the images useful, so based on the data and facts Google implemented this change to their algorithm. Another change the company is considering is to personalize search results more than a user’s location or immediate context from prior results. However, Google has escaped the criticism that Facebook and Twitter receive for creating a filter bubble where they are shown items they are predisposed to like. At the end of the meeting, CNBC found that Google takes a very analytical approach when ranking pages in their searches which do not operate on bias.
One way that Google decides which pages to show the user is PageRank. PageRank orders the webpages according to how many others link to it. If you consider how we update page ranks for a simple model, each node divides its PR over its outgoing links and a new PR is the sum of all values associated with inward pointing links. Google is responsible for processing billions of pages; thus, you can infer that outgoing arrows only hold small values. Conversantly, because hundreds of inward pointing arrows connect to a webpage, these minuscule values add up. This is the general idea for their PageRank algorithm which is way more complex in actuality. It would be easy to hack the system and change the PageRank of a site by creating false pages that link to it. So, Google is forced to update their algorithm periodically to prevent people from gaining the system and making a certain page more relevant. Using their “quality raters” also plays a big part to debug their algorithm and catch mistakes the computers would almost never pick up. An example of a screw up was having a white supremacist website appear as the highest-ranking website for the search “Did the Holocaust Happen?”. Google does realize their algorithm is not perfect, however with constant updates to it they can improve user satisfaction for an inquiry and keep out hackers.