With discourse increasingly migrating to digital spaces, where the visibility of topics is heavily dependent on post popularity, understanding the dynamics of online popularity has become crucial for both individual users and society at large. This thesis examines the mechanisms underlying post popularity on Hacker News, a social news platform known for its tech-savvy and discussion-oriented community. While traditional communication research on this topic is guided by theoretical frameworks such as news values or uses-and-gratifications theory, this study adopts a holistic, data-driven perspective leveraging predictive modeling.
For this purpose, a large-scale dataset of Hacker News submissions was independently collected and processed. The final dataset, consisting of a wide range of variables precisely describing each post, was then used to train a Random Forest Regressor. Subsequently, feature importance methods such as SHAP and Permutation Importance were applied to examine the captured relationships between characteristics such as user attributes, post content, and timing in relation to post popularity.
The analysis reveals that user experience plays a critical role, alongside contextual factors such as platform traffic patterns and the popularity of competing posts. Furthermore, frequent posting behavior is associated with lower popularity. Political content remains relevant despite platform restrictions, while well-crafted scientific and technical posts are associated with profounder success. Overall, Hacker News appears to reward high-quality, substantive contributions, suggesting that meaningful discourse remains achievable on social news platforms.