A study published in Nature found that Google search engine doesn’t generate biased results, but politically polarized individuals tend to click on links from partisan news websites, leading to self-segregation.
This suggests that it’s possible to escape online echo chambers, depending on user choices. Algorithms play a role in shaping our information consumption and opinions, but quantifying their impact on political polarization is difficult.
Katherine Ognyanova, a co-author of the study and a researcher at Rutgers University, recognizes that algorithms influence how we absorb information and form our judgments. However, determining the specific impact of algorithms on political polarization is difficult. According to Ognyanova, algorithms take into account a variety of parameters such as user identification, location, device, geography, and language, but the algorithm’s inner workings remain a mystery “black box.”
Ognyanova and her co-authors collected data by giving participants a browser plugin that captured their Google search results and the links they clicked on. They gathered information from hundreds of Google users in the run-up to the 2018 midterm elections in the United States and the 2020 presidential election in the United States.
The data was evaluated in relation to the age of the participants and their self-reported political attitude. This combination of real-world behavioral data and survey findings demonstrated that tailored news pieces based on political inclinations did not appear prominently in Google search results. The algorithm provided minimal customisation, and more partisan individuals were more likely to click on links that supported their pre-existing views.
Google responded by praising the researchers’ efforts and emphasizing its commitment to keeping a relevant and reliable system. It asserted that its search feature does not seek to infer sensitive information in its results, such as race, religion, or political affiliation.
The sources for this piece include an article in ScientificAmerican.