Predictive Analysis of Social Media Personalization and Its Effects on User Attention and Isolation.

ABSTRACT In this paper we examine the relationship between users’ social media usage behavior and their self-reported cognitive and social well-being. We used data from a public survey of 447 users to examine their average hours and aimless use of social media to assess time spent on social media and their levels of distraction, difficulty concentrating, and depression (as a proxy for isolation). Statistical analyses included Pearson correlations and a machine-learning feature importance model using artificial intelligence (Random Forest Regression) and visual comparisons of the distributions between key variables. We found a significant positive relationship between aimless scrolling (feed_similarity), for example, and feeling isolated. This result is in line with prior work on the attention economy and digital well-being, and suggests that high levels of algorithmic-driven engagement may be associated with negative cognitive and social outcomes.