Much has been made in social media and political class about potential biases within the algorithims of commercial internet search engines. At this point in history, anyone with a smart phone now has internet access nearly anywhere signal is available. Given the ubiquity of the internet, the idea that everyone has the Library of Alexandria within a device that can fit within a shirt pocket is no small feat for humanity—but how does one search through mountains of information, and misinformation? Enter the search engine.
This article attempts to examine the question of ideological biases within commercial internet search engines, and do so in as academic, and objective a manner as possible. Given the platform is a standard internet blog, it is understood this format may be offputting to some, perhaps even arrogant to others. Objectivity, however is the goal, thus the format.
If commercial internet search engines frame results designed to suit a particular ideology, then the results of identical controversial statements between various internet search engines will fit a pattern for each internet search engine provider, in an observable manner.
On 6 September 2018 GovPredict published a review of known political donations made by Alphabet Inc. This corporation is the parent company of Google, the largest search engine by an overwhelming margin. They concluded what many assumed: 90% of Alphabet’s employees that made a political contribution, did so to a Democrat candidate, or to an organization typically identified as beign sympathetic to the Democrat Party. Given GovPredict can be accused of being merely a review by a small, uncredentialed blog: in 2011 CBS News reported similar findings about Google’ political contributions.
Because of this, the assumption is that engineers at Google will tune their algorithms in a manner to suit their biases, wittingly or unwittingly. This is hardly a new accusation made towards Google, as this article by Business Insider from 2014 suggests. This is an accusation often made by right of center political groups. Who claim information presented by Google does not incorporate right of center interpretation of current events, ideas, and even basic facts that provide evidence of the merits of their ideas. The search results are designed to bury information that may lead a neutral observer to conclude in a manner consistent with left of center biases.
Interestingly, this accusation was presented as having merit by The Guardian on 6 September 2018, USA Today on 10 September 2018, and even previously by Slate on 7 December 2015. While USA Today can be considered politically moderate in it’s content, neither Slate nor The Guardian are publications considered to be right of center.
On 4 December 2018, a competing internet search engine, DuckDuckGo, explained how Google’s search algorithms can influence the presented search results by what they refer to as a filter bubble:
Put simply, it’s the manipulation of your search results based on your personal data. In practice this means links are moved up or down or added to your Google search results, necessitating the filtering of other search results altogether. These editorialized results are informed by the personal information Google has on you (like your search, browsing, and purchase history), and puts you in a bubble based on what Google’s algorithms think you’re most likely to click on.
The filter bubble is particularly pernicious when searching for political topics. That’s because undecided and inquisitive voters turn to search engines to conduct basic research on candidates and issues in the critical time when they are forming their opinions on them. If they’re getting information that is swayed to one side because of their personal filter bubbles, then this can have a significant effect on political outcomes in aggregate.
In simpler terms, Google does not present search results to suit their biases; the search results are intended to produce results that suit the user’s bias. If one never seeks opinions that differ from his or her own, one will never understand any one political issue beyond their own bias. This can lead to user’s simply viewing interpretations of current events, ideas, and even basic facts that provide evidence of the merit of their ideas, that only confirm their own opinions.
This study by DuckDuckGo presents findings that appear to correspond to one conducted by The Wall Street Journal during the 2012 presidential election. Here it was observed personalized results were provided for serch queries including the name Obama but not those with the name Romney. Google did provide an explanation why this was the case, and cited the number of searches queries that included Obama simply outnumbered those that included Romney. Personalized results may not be available for the latter due to lack of context in previous searches.
In the interest of full disclosure, the research for this Literature Review, was done with the assistance of the DuckDuckGo internet search engine.
A small number of subjects volunteered to search identical terms in three internet search engines. The three search engines chosen for this review:
The group of volunteers include the author of this article, with a total number of 7. To act as a control for individual biases between the group of volunteers, all of the volunteers for this study identify themselves politically as classical liberals, or in modern parlance, libertarians. Why libertarians? Libertarianism as a philosophy is neither right nor is it left. It is centered on recognition of individual rights. Often where libertarians agree on certain issues with the political right or the political left, it is from the viewpoint of the guarantee of individual rights rather than the fickle political justifications of the day. While choosing a group of libertarians specifically may imply bias towards libertarian leaning search results, the nature of the philosphy transcending both sides of the political divide is indeed a control.
Because any individual classical liberal/libertarian may have particular preferences towards where they find informaton on the internet and what search engine they use, another control in the search queries was added. None of the search queries are political in nature, however all of them are controversial. The following five statements were searched between the aforementined search engines:
- Deep dish pizza is not pizza
- The Beatles are overrated
- Butt implants are fake butts for fake people
- Coke is better than Pepsi
- Bolivian Air Force pilots cannot avoid mountains
Each volunteer was provided with a Microsoft Excel spreadsheet, and simply asked to copy and paste the top five search results that are not advertisements for each of the above five queries. All returned their completed spreadsheets by 14 December 2018.
Individual search results were identified with the first character in the internet search provider’s name (i.e. G is for Google), and by a numeral. The numeral is intended to correspond to an individual search result, therefore G1, D1, and B1 for example, are the same search result identified on all three internet search engines. It is in this way, a unique result can be identified should a particular search engine produce a unique result.
Due to the convenience of the population size, further analysis on how these results are presented in order for each user will also be observed.
The following are the results in tabular form:
When put in tabluar form, one can see the results for many of the queries are similar. Where they differ however is the order they are presented. One issue with the methodology is the limited scope of the results recorded, it is possible the search results are more or less the same when etended to the first page of results and beyond.
A noticable feature of the results, is the Google results cover a smaller spread. For example, for the first query, Deep Dish Pizza is not Pizza, Google only covers results 1-8 between seven people. The other two search engines however, cover a spread of 1-12. More concerning, are the order of results are nearly identical across all users.
Another thing that can be noticed is in the fourth query, Butt Implants are fake Butts for Fake People. If one were investegating this subject on Google, he or she would need to wait untiil the third search result (47) to find a search result corresponding with either of the other search engines. It is obvious this is a meaningless subject, however given the limited attention span of the average American for a subject more meaningful the third result can be significant. If one wants to question the result—so what if Google appears to have identical results between users on a search related to prothetic devices for the human posterior? The better question is, what if Google has identical results between users for a subject that actually matters?
While this on the surface it might appear DuckDuckGo’s claims have some merit; one can see their results cover a wider spread between users and much greater variance in the order presented when compared to Google. That said, the results are not all that different. In some cases one can see the same “filter bubble” DuckDuckGo accuses Google of presenting to its customers, within DuckDuckGo’s results. It would therefore appear the search engines do indeed present an observable pattern in the results. What that pattern is, if it can be considered a bias, and how it affects the user is not something that can be quantified here.