Google search results for immigration topics tend to present an overall positive view of the issue, according to a recent analysis of Swedish search data. The research, published in Heliyon, found that while the links skew slightly positive, the search engine tends to place neutral, unpoliticized sources at the very top of the page.
Search engines curate the information people see about social issues, raising questions about potential political bias. Some public figures claim that Google favors left-leaning content, which could theoretically influence voter behavior. This curation could also create filter bubbles where users are isolated from viewpoints that challenge their existing beliefs.
Prior scientific studies on this topic have been relatively scarce. Some previous research has suggested that Google might favor liberal media sources. Those studies often relied on indirect measures like how frequently articles were shared on social media by registered voters.
To investigate whether this bias actually exists in the text of the search results, researchers designed an experiment relying on direct human evaluation. Erik J. Olsson, a researcher at Lund University, led the study. He collaborated with Guy Madison of Umeå University and Axel G. Ekström of the Royal Institute of Technology in Sweden.
The research team selected eight search terms related to immigration based on prominent Swedish political debates. These translated phrases included “diversity,” “integration,” “immigration,” and “Islam.” They also included the phrase “crime and ethnicity,” an intensely debated topic in Swedish politics regarding statistical relationships between immigrant status and crime rates.
To test the search engine independently of individual user profiles, the researchers deactivated their search histories and wiped their web browser cookies. They then performed these searches from eight different municipalities in southern Sweden on the same day in November 2020. These locations were chosen because they are governed by political parties with contrasting stances on immigration.
The local leadership in these municipalities ranged from the immigration-friendly Social Democrats to the socially conservative Sweden Democrats. The researchers wanted to see if the search engine would alter the results based on the political leanings of the geographic area.
For each search term across every municipality, the team collected the top 30 web links provided by Google, excluding video results. This process resulted in a pool of 264 unique web pages. The researchers then hired a small group of five university students to read the content of each web page and rate its slant.
The raters evaluated whether the text on each page presented a negative, neutral, or positive attitude toward immigration. They also rated the level of politicization on a three-point scale. A neutral information source like an online encyclopedia would score the lowest, while an opinionated media piece would score the highest.
The study revealed an overall positive attitude toward immigration across the analyzed search results. However, the specific search phrase dictated the results heavily. Searches for the words “diversity” and “integration” returned highly positive web pages.
In contrast, searches for “Islam” returned completely neutral, educational links, such as study guides for elementary religion classes. The search phrase “crime and ethnicity” yielded more mixed and highly politicized results.
The researchers suspect that Google might heavily curate the results for sensitive topics like religious belief to avoid stirring controversy. For less sensitive search terms, the engine might generate results more organically, allowing for greater political diversity in the links.
The researchers also examined how Google ranked these pages. They found that the very first few links on a search results page were almost always neutral and unpoliticized. These top spots frequently directed users to government websites or established encyclopedias like Wikipedia.
The more opinionated and politicized links tended to appear lower down on the page. The same was true for links displaying a positive attitude toward immigration. Because the search engine placed neutral links at the top, its ranking algorithm actually reduced the overall positive slant of the results page compared to an unranked list.
A user would have to scroll past the top results to reach the more biased content. The political orientation of the search location did not alter the information provided by the search engine. Users searching from a conservative municipality received the exact same links as those searching from a liberal municipality.
The pages occasionally appeared in a slightly different order, but there was no location-based filter bubble effect. The researchers noted a few limitations in their experimental design. The study relied on a small group of young university students to rate the web pages.
A person’s own background can influence how they interpret text. This means a different demographic group might categorize the political leaning of these links differently. The search terms were also chosen by the researchers rather than generated by real users.
In everyday life, people with varying political views likely type different phrases into search engines. A conservative user and a liberal user might ask different questions about immigration, leading them to entirely different sets of information. Research shows that internet users tend to focus their attention almost exclusively on the top few links of a search page.
Because the top links in this study were mostly neutral, users are not immediately forced into a partisan filter bubble. Still, because Google provides a mix of positive, negative, and politicized links further down the page, users might construct their own bubbles by seeking out specific content. People often click on links that confirm their existing beliefs, allowing an individual to scroll past neutral top links to find an opinion piece that suits their worldview.
The study, “Is Google liberal on immigration? Attitude bias, politicisation and filter bubbles in search engine result pages,” was authored by Erik J. Olsson, Guy Madison, and Axel G. Ekström.