What’s in your PIE?

(Summary by Katherine Furl) 

 

What can social media users’ personalized information environments tell us about whether social media content is politicized and trustworthy? In “What’s in your PIE? Understanding the contents of personalized information environments with PIEGraph,” Deen Freelon, Meredith Pruden, Daniel Malmer, Qunfang Wu, Yiping Zia, Daniel Johnsons, Emily Chen, and Andrew Crist showcase the capabilities of PIEGraph, a software system they developed, to provide unique insight into the kinds of content users encounter on the platform.  

 

Freelon and coauthors investigate social media users’ personalized information environments, or PIEs. These unique environments—for example, users’ social media feeds—are constructed through users’ expressed preferences and predictions from algorithms. As PIEs become increasingly integral to social media users’ everyday lives, researchers, policymakers, and social media users broadly have become concerned about information quality in these spaces. 

 

To map out users’ PIEs, Freelon and coauthors developed the PIEGraph software system, and combined large-scale anonymized data collection from approximately 1,000 X/Twitter users with survey data detailing important information about participants’ demographics, mainstream media consumption, political leanings, and conspiracy belief. In doing so, Freelon and coauthors adopt a user-centric approach. In contrast to platform-centric approaches, which extract data directly from social media platforms, user-centric approaches instead employ user-interactive methods to approximate users’ PIEs. 

 

Freelon and coauthors’ novel approach uncovered some unforeseen trends in the PIEs of X/Twitter users. For example, Freelon and coauthors note that while “previous user-centric studies have found that 2%–3% of content that participants view is political,” over 10% of hyperlinks in this study were politically relevant. Additionally, respondents reporting more mainstream media consumption had more political, right-wing, and lower-quality content in their X/Twitter PIEs—and unlike past research, respondents’ age did not play much of a role here. 

 

Ultimately, Freelon and coauthors’ findings illustrate the need to track the potential risk of systemic exposure to low-quality content on platforms like X/Twitter over long periods of time. The authors stress government policies must protect researchers’ ability to map out these risks—without these protections, understanding the informational health of social media users’ PIEs becomes unnecessarily difficult.