Garrett, R. K. (2009). Echo chambers online?: Politically motivated selective exposure among Internet news users. Journal of Computer-Mediated Communication, 14(2), 265-285.
Summary:
The Internet offers users abundant access to information from around the world, which expose people to more diverse viewpoints, including both similar perspective and attitude-challenging information. With so much information, voice comes up that whether the Internet leads people to adopt increasingly insular political news exposure practices, i.e., “echo chambers” and “filter bubbles”, which refer to cases that individuals tend to be exposed only to information from like-minded individuals or materials similar to previous reading due to the influence of algorithms the news sites employed. This paper tests the viewpoint by conducting a web-administered behavior-tracking study. They find that the effect of opinion-challenging information is small and only marginally significant, which means worry that the Internet will lead to an increasingly fragmented society appears to have been overstated.
Reflection:
The paper shows readers how to conduct an exhausted and comprehensive experimental design when studying a specific problem, from selecting the research subjects to designing the experimental setting/workflow, e.g., using a purpose-built software administrates the study. Appropriate groups should be chosen to study according to the problem itself — two partisan online news services, left AlterNet and right WorldNet Daily, are able to bring immediate benefits to the study with the inherent engagement in selective exposure. Another thing I am concerned about is that there are not enough samples participating in the experiment, i.e., a total of 727 subjects (358 Alternet readers and 369 WorldNetDaily readers respectively), which possibly means the generalization of our observation result is not universal enough. Although the author mentions in the limitation that “when viewing these results is that the sample is not representative of the U.S. population at large”, I think it is very difficult to represent groups with hundreds of subjects, not to mention specific group like U.S. population at large. Why do not we find some volunteers on university campuses or research institutes, e.g., students and faculties, which is easier to get more volunteers?
Bakshy, E., Messing, S., & Adamic, L. A. (2015). Exposure to ideologically diverse news and opinion on Facebook. Science, 348(6239), 1130-1132.
Summary:
Based on the same problem background as the first paper, e.g., “echo chambers” and “filter bubbles”, the paper attempts to explore how this modern online Internet environment will influence people’s exposure to diverse content. Past empirical attempts turn out to be limited because of difficulties in measurement and show mixed results. To overcome previous limitations, they use a large, comprehensive Facebook dataset to measure the extent of ideological homophily and content heterogeneity in friend networks. They found that individual choice plays a stronger role than algorithmic ranking does in content exposure.
Reflection:
One heuristic point to take away is to classify stories as either “hard” (such as national news, politics, or world affairs) or “soft” content (such as sports, entertainment, or travel) by training a support vector machine, which reduces the data size while keeping the most efficient part of the dataset since “hard” content tends to be more controversial than soft content. However, some of the “soft” content is also practicable and appropriate for the task, i.e., hot and controversial topics in sports and entertainment fields. Thus the author may consider augmenting the size of the dataset to find out the result.
Same as the vague distinction between exposure and consumption mentioned in the limitation analysis, I also think how to clarify the line among conservative, neutral, liberal? Can we possibly give a uniform and appropriate criteria for the classification of those three categories?
