Bad actor reporting has recently grown in popularity as an effective method for social media attacks and harassment, but many mitigation strategies have yet to be investigated. In this study, we created a simulated social media environment of 500,000 users, and let those users create and review a number of posts. We then created four different post-removal algorithms to analyze the simulation, each algorithm building on previous ones, and evaluated them based on their accuracy and effectiveness at removing malicious posts. This thesis work concludes that a trust-reward structure within user report systems is the most effective strategy for removing malicious content while minimizing the removal of genuine content. This thesis also discusses how the structure can be further enhanced to accommodate real-world data and provide a viable solution for reducing bad actor online activity as a whole.
- The Algorithmic Reduction of Bad-Actor Reporting on Social Media