The Impact of Social Media Algorithms on Public Opinion

Social media recommendation systems play an important role in deciding what information users see and how often they see it. These systems use ranking algorithms, personalisation, and user engagement patterns to recommend content based on individual interests. Previous research has shown that these mechanisms can influence the type of information people consume, contribute to the spread of misinformation, shape political opinions, affect consumer decisions, and impact users' well-being. However, these effects are often examined separately, and simply showing users certain content does not necessarily mean that their opinions will change over time. To address this issue, this paper proposes a multi-objective machine learning framework that evaluates social media recommendations using a Civic-Impact Balanced Recommendation Score (CIBRS). The proposed score considers several factors, including content relevance, information quality, diversity of viewpoints, misinformation risk, and potential effects on user well-being. Based on these factors, a constrained re-ranking approach is introduced to recommend useful content while reducing repeated exposure to similar opinions and limiting the promotion of potentially harmful information. This research does not present new experimental findings from individual participants. Instead, the numerical charts are based on values reported in existing research studies, while the proposed CIBRS framework is presented as a new concept that requires further testing and validation. The overall analysis suggests that the influence of recommendation algorithms is more clearly visible in content exposure and user engagement than in long-term changes in political opinions. The main contribution of this study is a practical framework for evaluating and improving social media recommendation systems. Rather than focusing only on increasing user engagement, the proposed appro