User Voices, Platform Choices: Social Media Policy Puzzle with Decentralization Salt

In the current digital era, social media platforms wield crucial influence, with the potential for biased content moderation. Considering this risk, we propose a decentralized social media policy-making in this work. The noticeable difference between people's preferences and X's established policies in a preliminary study motivates us to design a similar website to collect more comprehensive data in a diverse community. Consequently, N=110 individuals from diverse backgrounds participated in our primary experiment to decide about content moderation on social media. For this purpose, 546 tweets in 3 categories are investigated, 3032 records are captured, and the effect of personal favor on content moderation is analyzed. Furthermore, we propose a novel AI-based method to learn the recommended policy of participants and achieve an accuracy of 79%. Also, by considering the suggested policy of 5 Large Language Models, it is illustrated that they cannot be the decision-makers on democratic social media platforms.