In a recent article right here on SEW, Christopher investigated the manipulation of UK political party autosuggest queries, but what exactly is search engine bias?
Search engine bias is the unfair skewing of results and it sits within the intermediary mechanism of search systems (see diagram below) and biases can be split into three parts:
1) Non-neutral search engine technology
Years ago studies focused exclusively on search software. Soon people realised that the user interface is of equal importance to any algorithm, as this is the view we get of the system and our successful use of it.
Now, this is where it gets even more interesting, the presentation of results influences free-thought which powers interaction. As a result, all search engines are bias because of the design and user interface.
We have two types of result inspection bias: quality and trust.
Quality bias means that the searcher will subconsciously take into account surrounding results and if they’re not relevant our behaviour can change, we may decide to retype our query or inspect a site we otherwise would not have dreamt of.
This is precisely the reason why engines have quality scores for their paid search systems, if paid results are irrelevant this can mess with organic result inspection.
For the most part search algorithms are generally very effective at presenting us with relevant results so we trust the engine, this is known as trust bias.
Personalisation through cookies and the creation of 3D cubes are used to increase relevancy.
While there are many advantages of personalisation, personalisation restricts results and is a form of search engine bias.
2) Result manipulation
Political and social bias came to the fore in 2002 when Google was temporarily blocked in China and since then, with the USA’s Digital Millennium Copyright Act, indexes and results have always been filtered to comply a little better with the law.
Let’s now take this the other way around, is search engine coverage biased along national lines?
In short, yes. Engines do not have proportional coverage of a particular group of websites or regions. But why is this? Well, country population differences and internet penetration cause unintentional bias.
Bloggers make up PageRank, a two-fold mathematical and human-centric algorithm. If a country has more bloggers they are more likely to link to fellow blogs within that country. Throw in personalisation, in which geographic location is taken into account, this further reinforces country specific blogger-linked networks.
This is why .com ccTLD’s rank more than other ccTLD’s and why many marketers and business people alike want .com domains over other ccTLD’s. This is a classic example of success-breeds-success whereby the US has the upper-hand due to the fact of it being an early adopter of the web.
Search engines may think of themselves as being objective but like any other media company, editorial judgements are made and are factored into automated operations. Engines trust certain sources more than others.
On Google, Wikipedia has been regarded as a trustworthy source. This may be because Wikipedia contains up-to-date, current and reliable information that answers a question through well-researched, referenced, thus reliable, content. As a result it often features in the top results.
If the top results do not satisfy the searcher, the search is unsuccessful. Since each unsuccessful search diminishes search engine perception, search engine bias is often towards trustworthy sources.
This is why Wikipedia will not disappear anytime soon and it also explains why Google’s quick answer boxes are at the top of results – to answer a query even quicker and to increase an engine’s performance perception.
So there you have it, our results are skewed because of search engine bias.
Biases singlehandedly have the biggest impact on results because they encompass user interface and influence over result inspection; personalisation; and finally ranking algorithms (which is part mathematical part human edited).
Please note: this particular article took into account bias from an index perspective. Another important angle to take is from a keyword perspective. Which search engines are pro-life if you type in euthanasia, for example?
Further classic studies:
- Information Retrieval from the Internet: an Evaluation of the Tools
- Search Engine Bias and the Demise of Search Engine Utopianism
- Search Engine Coverage Bias: Evidence and Possible Causes