| Popis: |
Online social media platforms significantly influence public debates by shaping the information users encounter. Content visibility on these platforms is regulated by recommendation algorithms designed to maximize user engagement using individual-level data, including personal preferences and interactions. These algorithms play a crucial role in information dissemination, yet their inner workings are often undisclosed, raising concerns about potential manipulation of visibility. While algorithms may be intended to limit the spread of harmful content, they can also be exploited to suppress dissenting voices without users' awareness. The suspicion that platforms deliberately reduce the visibility of certain users or content - commonly known as shadow banning - has garnered significant public attention, with numerous figures advocating for greater transparency around this practice. In this study, we perform a quantitative study geared to identify suspicious changes in content visibility on Twitter (now known as X). We build and study a dataset of over 13 million tweets from more than 5 million users discussing the Ukraine conflict, including each tweet's number of views and metadata, aiming to detect reduced or inflated visibility patterns. We investigate how visibility correlates with factors such as authors' stance, role, interaction networks, and content narratives. Our findings reveal significant variations in visibility, likely driven by algorithmic interventions. These results highlight the need for greater transparency in regulating online information ecosystems to prevent algorithmic manipulation that could undermine public discourse and the fairness of debates. |
