Most of us rely on Google to explore the internet, look up recipes, shop for gifts, read the news, and even research political candidates. It is truly the king of search engines and Google has gone to great lengths to stay at the top of its industry. Continued adjustments to their search algorithm ensure that users see useful results that are likely to be clicked on. However, it seems that this search engine has a downside. It turns out that the search results we see might be influencing our opinions on important topics.
San Diego, California middle schooler, Agastya Sridharan, first became concerned about Google’s impact on the public after reading an article in the Wall Street Journal about bias in Google search results. At the time, Agastya, 13, decided to dedicate his science project to finding some answers. He wanted to know if the order in which search results appear can alter a person’s political opinion. The results of his experiment are intriguing and thought-provoking, especially as the American political landscape continues to be divided.
How did he do it?
Agastya had a difficult task on his hands with conceptualizing and mobilizing a project involving political opinions right before the 2020 presidential election. He had to find a way to get non-biased data from groups of volunteers during one of the most politically charged periods in recent history. To accomplish this, he came up with a two-survey system that asked people to choose between two candidates after reading their biographical information. Afterward, the volunteers were shown pages of search results that compared the candidates. The idea was to find out if the order of the search results would change their opinion of the candidates.
In the first survey, he used two fictional candidates — Ronald Bush and Julia Hillard — that were based on Donald Trump and Kamala Harris. After a thorough interview in which 100 volunteers gave some demographic information, they were asked to read biographical information on each candidate and asked to form an opinion on who they would vote for in an upcoming election. Afterward, each volunteer was shown Google search results for the phrase “Julia Hillard vs Ronald Bush.” The volunteers then were allowed to click on and read as many of the results as they wanted and state whether or not this information had changed their vote.
The search results were simulated using HTML to give the appearance of an actual Google search. Each article was taken from allsides.com and altered only to show changes in the biographical information (e.g. changing the name Donald Trump to Ronald Bush etc.). Agastya chose the source, allsides.com, because it is an unbiased, credible, and trusted source that rates articles based on factors such as the author, news publication, and community ratings. Using this information, the articles were carefully chosen to reflect liberal, conservative, and non-partisan viewpoints.
A total of 25 search results were used, five of which were rated as strongly liberal, five were rated as moderately liberal, five were rated as non-partisan, five were rated as moderately conservative, and five were rated as strongly conservative. These results were then organized into five different sets. Each set ordered the results according to a different political bias. The first set put the liberal results first, the second put the conservative results first, and the third mixed liberal and conservative results equally. The fourth and fifth sets did something a little different — they took the liberal and conservative sets and switched the third and twenty-third results. The intention was to increase credibility by including a competing viewpoint. Each volunteer was shown only one of these sets.
The second survey used the same method but instead of the fictional candidates, the names Donald Trump and Joseph Biden were used. Also, Agastya was able to use Amazon’s Mechanical Turk, a crowdsourcing marketplace, to obtain 250 paid subjects that were more geographically diverse.
After reading through the search results the volunteers and subjects were asked if they would like to change their vote. If they changed their vote in the direction of the search result bias, they were counted as having been influenced by the search results. To put it another way, if people were shown liberal search results first and then changed their vote to the liberal candidate they were counted as having been being influenced.
The graphic below showcases Agastya’s results, and not only did they prove his hypothesis, but they did so with alarming numbers.
The results of the surveys are provocative and shocking. It would seem that the order of Google search results has a heavy influence on political issues. The results of the fourth and fifth sets are particularly interesting as they had a much stronger influence on the readers, which means Agastya’s goal of swapping results to achieve added credibility has worked.
It has been said that Google’s algorithm is geared toward showing results that will get clicked on; the site intentionally shows us what we want to see. When searching for recipes or things to buy this can be ideal, however, in the American political arena maybe it’s doing more harm than good. It seems that by giving us what we want Google has contributed to confirmation bias.
Agastya’s project sheds some light on a topic that isn’t discussed much, but it may be time to start thinking about it. We asked Agastya about the project results and he had this to say: “I think now, more than ever, we need to reflect on the current state of affairs in our nation. In a recent interview, Barack Obama called the internet the gravest threat to our nation, and I agree. What we’re seeing throughout the globe is the general decline of democracy and the rise of populism, evidenced by recent events in Brazil, Australia, and even within our own borders. I think a large part of this is the loss of shared reality. We can’t have meaningful discussions and disagreements if the entire country operates on a different state of facts.”
After completing his project Agastya applied to the Broadcom MASTERS competition where he became a finalist and won the Broadcom Leadership Award. The winner of this award is chosen not by a panel of judges, but instead by the other students to speak on behalf of their class at the Award Ceremony.