Is Google politically biased?

Analysing the accusations against Google's algorithmic neutrality

9 minutes to read
Article
Tunde Farago
07/03/2019

More than a billion people turn to Google every day as their primary source of information. We hear it very often that if you don’t know the answer to a question, just “google it”. As it has become somewhat of a habit of ours to just “google it”, none of us really questions the results we get, let alone thinks that there is some hidden political agenda behind them. But even if there was one, can we really know that?

Algorithmic neutrality vs political bias

Back in August 2018 Trump accused Google’s search results of being biased and supporting Democratic ideals. In a series of tweets the US president alleged that their search engine is showing only “Fake News” when one googles “Trump News” while at the same time hiding articles from conservative media. (Helmore, 2018; Superville and Ortutay, 2018) Since then, several news outlets, including the Guardian, have raised the question whether or not Trump is right to accuse Google of being politically biased. (Hancock et al., 2018)

Following these accusations the CEO of Google, Sundar Pichai, appeared in front of the US Congress in December 2018 to defend the company’s search engine system. He explained that it is up to the algorithms that work behind the search engine to determine what gets to be shown as relevant, popular or trending based on objective guidelines. (Palmer-Derrien, 2018) However, not everybody was convinced by his explanation, and the question is: should we be?

When the president of the US calls into question Google's search engine, the whole reputation of the company and its authority in the field of knowledge production are at stake.

According to Gillespie, algorithms “are inert, meaningless machines until paired with databases on which to function.” (2014: p. 169) But when paired, they help us skim through the vast amounts of information on the Internet, picking out the hot and trendy things and at the same time hiding what they determine to be irrelevant information. (Ibid: p. 167) Behind Google’s search engine there is not one but several algorithms that work together to provide us with the best possible answers to our questions.

However, even before we start to search for something on Google, their web crawlers will tag and index all the websites and organize their content into the Search Index. Then the Google ranking system, called PageRank, comes along and chooses the most relevant information from these categories, examining at least 200 signals to provide us with the best suggestions. (Google Search, n.d. and Gillespie, 2014: p. 175)

Developers of search engines very often pride themselves in that the suggestions their tools provide are the results of algorithmic objectivity. (Gillespie, 2014: p. 179) “No provider has been more adamant about the neutrality of its algorithm than Google, which regularly responds to requests to alter its search results with the assertion that the algorithm must not be tampered with.” (Ibid: p. 180) Google’s 'About' page, which Gillespie calls their manifesto, suggests the same thing: “We never manipulate rankings to put our partners higher in our search results and no one can buy better PageRank. Our users trust our objectivity and no short-term gain could ever justify breaching that trust.” (Google, n.d.)

What should we make of the associations that algorithms claim to identify about us as a society — that we do not know, or perhaps do not want to know?

These claims, Gillespie argues, play a major role in Google establishing a monopoly over the production of knowledge online over its competitors. At the same time, defining the algorithms as objective and stating that they do not interfere with their workings, helps them fight off accusations of being biased, of creating mistakes and manipulating their search results. (Gillespie, 2014: p.185) So, when the president of the US calls into question Google's search engine, the whole reputation of the company and its authority in the field of knowledge production are at stake.

It was due to such accusations that the CEO of Google, Sundar Pichai, had to appear in front of the US Congress. While trying to describe how the algorithms work, Pichai also noted that “Google matches the keywords against web pages and ranks them based on more than 200 signals, including relevance, freshness and popularity.” (Palmer-Derrein, 2018) He also added that they do not “manually intervene on any particular search result” (Ibid.), once again claiming that it is up to their mechanically neutral algorithms to decide what gets to be picked up as the best suggestion.

Google's relevance criteria

The fact that when you google “Trump” the results will be dominated by allegedly left-wing political articles about him, does not necessarily mean that the search engine is picking on him or discriminating against him. It can also be the case that the algorithms simply determined that these articles were more 'relevant' for you than others. “When users click “Search,” or load their Facebook News Feed, or ask for recommendations from Netflix, algorithms must instantly and automatically identify which of the trillions of bits of information best meets the criteria at hand, and will best satisfy a specific user and his presumed aims.” (Gillespie, 2014: p. 175)

The idea is that the search engine can calculate what is relevant for the users, and what information to hide. Yet what is 'relevant' is not always clear. Relevant can mean that it is popular, matches the user's preferences, or is ranked as relevant due to other users' activities. The criteria by which relevance will be determined are embedded into the workings of the algorithms. (Ibid.)

Gillespie (2014) argues that knowing how the algorithms behind public search engines really work is tied to the very notion of power. Not everybody can get access to a company’s secrets. First of all, that is because they do not want the competition to get the upper hand and develop the same or better algorithms. Secondly, even if the designers would make the algorithms known to the public it would require major technical literacy from our side to understand them. (Ibid.: p. 185) Another potentially troubling result of revealing the algorithms' true workings could be that users might work their way around them or even manipulate them, thus rendering their criteria of relevance meaningless and invalidating the data they yield.

Yet we cannot deny the actual influence users have on the results of algorithms even when not knowing how they really work. Although technological neutrality is promised by the company that does not mean the story ends there. How the users use the service will also affect the said algorithm. (Gillespie, 2014: p. 182) According to Gillespie, “algorithms are made and remade in every instance of their use because every click, every query, changes the tool incrementally.” (Ibid: p. 173)

Users' influence on Google's search engine

Google's PageRank works by gathering information about its users' preferences when browsing the internet. Engaging with specific content adds weight to that website’s relevance index. (Gillespie, 2014: p. 178) Therefore, what gets to be shown on the first page of Google search results is not necessarily what we as individual users think it is the most relevant content for us to see, but rather in a sense reflects a wider public opinion. That being said, one might argue that Trump's accusations of Google being politically biased might have another explanation. It could be that lots of people who use Google search to get updates about world news and specifically American news prefer websites and media outlets that favor left-wing politics.

Gillespie (2014: p. 187) points out that “Google’s search engine, amid its 200 signals, does presume that relevant knowledge is assured largely by public ratification, adjusted to weigh heavily the opinions of those who are themselves publicly ratified. The idea that Google’s search results represent what the public already thinks is a bit dangerous. We should ask ourselves the question whose opinion matters in constructing Google’s PageRank? And who is left out?

Understanding these algorithms also sheds some light on what we as users think is important.

Gillespie also adds that there is an even more important question to ask: what kind of a public is being created this way, and whether or not this public would exist otherwise? (Ibid: p. 189) On the other hand, the data that algorithms use is collected from us, the users. Gillespie is quite right to ask then “what should we make of the associations that algorithms claim to identify about us as a society—that we do not know, or perhaps do not want to know?” (2014: p. 190)

Another reason the CEO of Google was being questioned by the US Congress was the allegation that if one searches with the word “idiot” in Google and clicks on images, the first recommendations one will get are pictures of Donald Trump. (Palmer-Derrein, 2018) Since we know that Google is really proud of their search engine being objective and democratic and tries to improve it as soon as allegations start to cumulate, one might think this will soon happen with the results of the word “idiot” as well. After all, Google already altered its search results for images when racist Photoshopped pictures of Michele Obama surfaced. (Gillespie, 2014: p. 182) Who says it will not do the same for Trump?

This also raises some concerns about freedom of speech. If Google will oblige and hide Trump as the answer for every negative query, will that then not be politically biased? The fact that the censorship will render negative search results about Trump invisible, will not stop certain users to think of him as an “idiot”, but it will be a serious threat to democracy and freedom of speech. Especially considering the fact that search engines are like media outlets and bending them to the wishes of politics will only render them useless and politically biased. (Gillespie, 2014: p. 168)

Objectivity trap and the risk of being biased

The aim of this article was to discuss the issue of political bias in relation to Google's search results algorithms. Considering that all the algorithms are built and maintained by engineers who have certain values, norms, prejudices and interests, and the fact that gaining full access to their true workings is something that we as average users will never have, might be enough to claim that there is some political bias behind the results after all. Yet that might be only part of the answer, since as we have already seen, many factors play a role in shaping and producing these results.

Understanding these algorithms also sheds some light on what we as users think is important. When it comes to Google’s ranking system deciding what is relevant and what to hide as irrelevant, our clicks and queries also matter. These processes after all reflect a more general public opinion. However, we should also be cautious to make such assumptions since it is not clear whose queries and activities are being considered as relevant and who is left out. Even though they might not be representing the actual public’s opinion, nevertheless, these results do say a lot about our values, norms and prejudices. The question is, is hiding them the solution?

Censoring everything that might be contradictory to our views and altering the algorithms every time one feels like it will not change things for the better. That will only endanger freedom of speech and therefore our democratic ideals as well. At the end in the efforts to make our search engines more objective, we would fall in the same trap of being politically biased.

References

Gillespie, T. (2014). The relevance of algorithms. In Gillespie, Tarleton, Pablo J. Boczkowski and Kirsten A. Foot (eds.) Media technologies: Essays on communication, materiality and society. MIT Scholarship Online: p. 167-193.

Google. (n.d.). Ten things we know to be true. Accessed on 18/12/2018.

Google Search. (n.d.). Our mission. Accessed on 19/12/2018.

Hancock, J., Metaxa-Kakavouli, D. and J. Park. (2018). Are Google search results politically biased?. Accessed on 18/12/2018.

Helmore, E. (2018). Trump accuses Google of promoting Obama's speeches over his. Accessed on 18/12/2018.

Palmer-Derrien, S. (2018). “The algorithm made us do it”: Google chief forced to defend search algorithm, as ‘idiot’ search brings up images of Donald Trump. Accessed on 18/12/2018.

Superville, D. and B. Ortutay. (2018). Trump accuses Google of biased searches, warns ‘be careful’. Accessed on 18/12/2018.