Google’s search algorithm changes?
https://www.cnbc.com/2018/09/17/google-tests-changes-to-its-search-algorithm-how-search-works.html
This article talks about the possible rigging of search results in terms of political bias, ever since President Trump accused Google of being biased. Though individual Google employees lean towards liberal on their political views, but there is no proof that search engine results are purposely skewed toward any direction. To “demystify” how Google runs its search engine, CNBC was invited to sit in on an internal meeting, and the article discusses the three big things they learned: How Google search works, How Google decides to tweak its search algorithm, and How Google experiments.
How Google search works: The search engine does not know what the input words mean, thus, it is merely looking for pages with those words, their synonyms, or common misspellings. Since it looks for relevant pages related to the word itself, it will not know if the word is used in a different context. Also, they learned that the company intentionally doesn’t reveal all factors used in the algorithm for ranking pages, partially due to prevention of people trying to manipulate information in the system, and also to stay ahead of potential competitors. This shows that this idea of networks is not just an algorithmic problem, but also a social issue as well.
How Google decides to tweak its algorithm: When considering a change in its search algorithm, Google tests the adjustment with a small percentage of real users and “search quality raters” to see the interactions. The raters are able to see the old and new results, being able to determine which is “better”. “Better” is defined by rating guidelines, such as a page’s expertise and trustworthiness. One important point made in the article is that they expressed how the meeting was not exciting. CNBC said that there were no passionate debate or philosophical concerns, as the data was the only factor driving the decisions.
How Google experiments: Google uses hard data to approve changes, but to choose the changes is a different and less disclosed process. Sometimes it’s from feedback, or sometimes, it comes from broad company directives/priorities. There is little to no search personalization, since it can cause people to lose trust in Google (as it creates filter bubbles).