The rise and continuous advancement of Information and Communication Technology, for example, social media in the past couple of decades have sparked intense interest among media scholars in comprehending their impact on shaping public opinion and its political ramifications. Currently, it is undeniable that social media-driven public opinion holds significant influence. Due to easy access to the internet and the fast spread of social media, we are more able than ever to track the opinions of individuals we have never met in person through online platforms. It seems more straightforward and efficient for governments and companies to detect public opinions than before. However, the intricate information dissemination and communication practices on social media present challenges in accurately predicting the formation of public opinion. Detecting public opinions towards the news is essential for politicians and companies to adjust their policies and strategies in time. Research about the sharing of information and viewpoints has raised concerns about the potential polarization and increasing political homogeneity of certain segments of the public attributable to social media.

This thesis project consists of four main sections. The initial section conducts a comprehensive literature review on information dissemination and the factors influencing opinion expression. The second section proposes a simulation-based framework to uncover the divergence between observed and actual public opinions, exploring different assumptions and metrics for opinion distribution. The third section employs the Beta distribution and modified Wasserstein distance to simulate opinion-sharing scenarios under various factors such as posting cost, rating mapping, bribes, and moral cost. The results indicate that these factors tend to generate greater divergence when true public opinions lean negatively but have a minor effect on negatively dominant true public opinions. The fourth section presents empirical findings based on the theoretical framework, suggesting that the developed approaches can effectively assess public sentiment related to homophily and polarization. These computational approaches can be automated, facilitating real-time transparency in forming public opinions on social media platforms.

This research contributes to the existing literature by proposing an approach to inferring the true distribution of public opinion. It also considers the impact of neglecting the silent majority's opinion on estimating true public opinions, which has yet to be explored. By examining the sensitivity of various factors to the divergence between observed and true public opinion distributions, this research provides a guideline for evaluating the effectiveness of current online review system designs. Furthermore, it benefits policymakers and product managers by enabling them to make better-informed policy adjustments by understanding true public opinions without relying on large-scale opinion polls.