US President Donald Trump last week accused Internet giant Google of abusing its monopoly over online search by “suppressing positive stories” about his administration. He complained on Twitter that “Google search results for “Trump News” shows only the viewing/reporting of Fake News Media. In other words, they have it RIGGED, for me & others, so that almost all stories & news is BAD”, and told White House reporters that Google had “really taken a lot of advantage of a lot of people, it’s a very serious thing”.
The allegations were refuted by everyone from search experts to critics of Google. What also came out, however, was that not a lot is known about how the search engine that is an inalienable part of almost everyone’s online life, works — and how it picks results. Google does not disclose much about its search formula because of two reasons: one, its entire business is based on it, and two, a lot of people would try to game the system if they got to know how it was done. So, if a certain article or page shows on top when you key in a certain search term — despite there being millions of pages on that topic for Google to scan through — it could be because of hundreds of factors, most of which are unknown.
The Google algorithm
And yet, certain things are known about search. The primary factor that determines the Google algorithm is the PageRank formula invented by its founders Larry Page and Sergey Brin over two decades ago. PageRank differentiated Google Search from what Yahoo and AltaVista was offering then, because it gave value to how many sites linked to a certain page. While this is just one of the many signals Google looks at, it’s still probably one of the most important ones.
There are many other factors that help a page to rank well — like the URL structure, headlines, bounce rate of the page, and the time spent on the page. There is a whole industry that is built around trying to game this system by peppering pages with keywords and packing in other elements that could help a page do better. But this so-called search engine optimisation (or SEO) does not always help, as other factors like authority of the website and its vintage cannot be faked. Over the years, Google has updated its algorithm at regular intervals to prevent people from taking advantage of it. Each update brings some pages up and others down. So in the case of Trump, if all the negative stories are being done by authoritative websites like The New York Times or CNN, and the positive stories are limited to the relatively small rightwing sites, then the former is likely to show up on top more often.
The human intervention
While Google has said its search is not affected by political ideology in any way, a report in The New York Times said that after a scrutiny over misinformation following the 2016 presidential election, Google had found that “0.25 per cent of its daily traffic linked to intentionally misleading, false or offensive information”. This is where it tweaked the algorithm to surface more “authoritative” content, and this led to complaints of a drop in traffic for many sites, the report said.
Last year, Ben Gomes, Google’s V-P for Engineering and Search, had told The Indian Express that whenever Google makes a change in the search algorithm, it asks 10,000-odd raters which option is better. “We ask them to judge based on rater guidelines, which is essentially a description of what search does. Using these raters, we change the algorithm, hopefully making them better and better,” Gomes had said. The raters’ feedback is used to flag issues with results, and not to affect the search algorithm.
But it may be wrong to think that Google Search cannot be influenced. In fact, as Google moves towards a search format where users don’t really have to search to find a result, it is using artificial intelligence and machine learning to preempt a query and surface a result. This means it knows enough about you to know what you will search, or the type of source you are likely to click more on. This could mean results get prioritised — and are hence not really neutral. So if a user is biased against Trump, there will be parts of the search, like the newsfeed, where he will see more of anti-Trump results. Also, suggestions often tend to suggest what people end up searching, and this again is not plain vanilla and is based on a lot of factors.