Bots on Social Media: The Past, Present and Future

Bots, and their more sophisticated variant, social bots, have become a ubiquitous part of social media. From a technological standpoint, the modern social bot is a remarkable achievement, having survived countless attempts at detection and removal by social networking sites through the evolution and persistence of those who operate bots. Social bots have been accused of being capable of influencing opinions and even manipulating election results, although recent work has also explored benign and benevolent use cases for social bots. This dissertation uses a variety of methods, from machine learning and network analysis to experiments, to study these different types of social bots.
The dissertation is based on five publications that contribute to our overall understanding of social bots and how to study them. The first two publications represent the early and naive era of social bot research, where the goal was to use machine learning to detect and study bots as manipulators of elections or spreaders of conspiracy theories. The third and fourth publications jump to the modern era of generative AI-powered social bots and focus on the bot detection capabilities of humans rather than machine learning models, an understudied area of social bot research. The fifth and final publication builds on the methods developed in the third paper and proposes a more generalized approach to using foundation models and generative AI in experiments to study phenomena such as social bots. Overall, this dissertation describes the history and evolution of both social bots and the field of study itself and concludes with an epilogue that speculates on the future of bot research in an era where Twitter is no longer a viable data source.