We’ve all come across examples of how Google can get things wrong: incorrect supermarket opening hours (http://www.rba.co.uk/wordpress/2015/01/02/google-gets-it-wrong-again/), false information and dubious sources used in Quick Answers (http://www.rba.co.uk/wordpress/2014/12/08/the-quality-of-googles-results-is-becoming-more-strained/), authors who die 400 years before they are born (http://googlesystem.blogspot.co.uk/2013/11/google-knowledge-graph-gets-confused.html), a photo of the actress Jane Seymour ending up in a carousel of Henry VIII’s wives (http://www.slate.com/blogs/future_tense/2013/09/23/google_henry_viii_wives_jane_seymour_reveals_search_engine_s_blind_spots.html) and many more. What is concerning is that in many cases no source is given. According to Search Engine Land (http://searchengineland.com/google-shows-source-credit-quick-answers-knowledge-graph-203293) Google doesn’t provide a source link when the information is basic factual data and can be found in many places. But what if the basic factual data is wrong? It is worrying enough that incorrect or poor quality information is being presented in the Quick Answers at the top of our results and in the Knowledge Graph to the right, but the rot could spread to the main results.
An article in New Scientist (http://www.newscientist.com/article/mg22530102.600-google-wants-to-rank-websites-based-on-facts-not-links.html) suggests that Google may be looking at significantly changing the way in which it ranks websites by counting the number of false facts in a source and ranking by “truthfulness”. The article cites a paper by Google employees that has appeared in arXiv (http://arxiv.org/abs/1502.03519) “Knowledge-Based Trust: Estimating the Trustworthiness of Web Sources”. It is heavy going so you may prefer to stick with just abstract:
“The quality of web sources has been traditionally evaluated using exogenous signals such as the hyperlink structure of the graph. We propose a new approach that relies on endogenous signals, namely, the correctness of factual information provided by the source. A source that has few false facts is considered to be trustworthy. The facts are automatically extracted from each source by information extraction methods commonly used to construct knowledge bases. We propose a way to distinguish errors made in the extraction process from factual errors in the web source per se, by using joint inference in a novel multi-layer probabilistic model. We call the trustworthiness score we computed Knowledge-Based Trust (KBT). On synthetic data, we show that our method can reliably compute the true trustworthiness levels of the sources. We then apply it to a database of 2.8B facts extracted from the web, and thereby estimate the trustworthiness of 119M webpages. Manual evaluation of a subset of the results confirms the effectiveness of the method.”
If this is implemented in some way, and based on Google’s track record so far, I dread to think how much more time we shall have to spend on assessing each and every source that appears in our results. It implies that if enough people repeat something on the web it will deemed to be true and trustworthy, and that pages containing contradictory information may fall down in the rankings. The former is of concern because it is so easy to spread and duplicate mis-information throughout the web and social media. The latter is of concern because a good scientific review on a topic will present all points of view and inevitably contain multiple examples of contradictory information. How will Google allow for that?
It will all end in tears – ours, not Google’s.