Trolling is a deviant behavior that provokes emotional responses, diffuses misinformation, or otherwise disrupts on-topic discussions. This study investigates trolling behavior as a complex dynamic process. The data comprised over 13 million Reddit comments, which were classified as troll or non-troll messages using the BURT model, fine-tuned with human coding set. By employing a unique, minimally complex, maximally predictive model from statistical mechanics and information theory, i.e., epsilon-machines and transducers, we find that engaging in trolling behaviors is both self- and other-motivated. Social inputs provide more information for predicting the future behaviors of the dynamic. Social inputs are also shown to encourage and reinforce trolling behaviors among Internet users. The article serves as a showcase for the use of information-theoretic measures from dynamical systems theory to conceptualize social media processes as computational algorithms executed by collectives of human users.