Auditing the audits

Evaluating methodologies for social media recommender system audits

Bibliographic Data

ID19039986
AuthorsPaul Bouchaud (0000-0002-0310-8194, corresponding author), Pedro Ramaciotti Morales (0000-0002-3649-4503), Pedro Ramaciotti
Year2024
Volume9
Issue1
Publication date2024-09-16
Peer ReviewedYes
Open AccessYes
TypeARTICLE
VenueApplied Network Science (JOURNAL)
Journal identifiersISSN: 2364-8228 • E-ISSN: 2364-8228
PublisherSpringer Nature (PUBLISHER • SG)
DOI10.1007/s41109-024-00668-6
OpenAlexW4402702536
LanguageEN
References cited39

Through a simulated Twitter-like platform designed to optimize user engagement and grounded in authentic behavioral data, this study evaluates methodologies for auditing social media recommender systems. Our analysis focuses on the impact of key parameters in sock-puppet audits, the number of friends and session length, on audit outcomes. Additionally, we investigate the algorithmic amplification of political content across different levels of granularity, segmenting users based on political leanings and considering multiple political dimensions beyond declared affiliations. Our findings underscore the necessity of employing realistic parameter settings in audits and highlight the importance of nuanced political segmentation. Amid increasing regulatory scrutiny, this research contributes to enhancing methodologies for auditing social media platforms

Audit · Business · Information retrieval · Internal audit · Joint audit · Operational auditing · Recommender system · Social media · World Wide Web · Complex Network Analysis Techniques · Computer Science · Misinformation and Its Impacts · Social Media and Politics · Accounting

Tools

Open DOIOpen Access