With the changing environment, it’s essential to study the information consumption of social media users and to audit how automated algorithms (like search and recommendation systems) are modifying the information that social media users consume. In this thesis, we fulfill this high-level goal with a two-fold approach. First, we propose the concept of information diets as the composition of information produced or consumed. Next, we quantify the diversity and bias in the information diets that social media users consume via the three main consumption channels on social media platforms: (a) word of mouth channels that users curate for themselves by creating social links, (b) recommendations that platform providers give to the users, and (c) search systems that users use to find interesting information on these platforms. We measure the information diets of social media users along three different dimensions of topics, geographic sources, and political perspectives.
Our work is aimed at making social media users aware of the potential biases in their consumed diets, and at encouraging the development of novel mechanisms for mitigating the effects of these biases.