Simulating Reputation Dynamics and Their Manipulation – An Agent Based Model Framework

Abstract

Reputation is essential to human interactions and shapes group dy-namics, however, it can be manipulated. In order to identify key aspects of malicious communication strategies, we have developed an agent-based simulation framework that captures aspects of the dynamics of social reputation communication: the reputation game simulation. After giving an overview of our framework, we highlight both previous and new results obtained with it. Similarly to other works in the literature on trust and reputation networks, probability functions and Bayesian logic are used in the reputation game simulation to represent uncertain-ties in agents’ beliefs. A new aspect of our framework is how bounded rationality of humans is modeled. It is regarded as a consequence of the necessary data compression step minds with limited capacity have to perform. Although this tries to minimize the loss of relevant infor-mation, for which we discuss two theoretically plausible options, it introduces cognitive imperfections. The resulting imperfect reasoning due to this and other cognitive shortcomings makes agents vulnerable to deception. This eventually leads to the emergence of communication and behavioral patterns in reputation game simulations that resemble reality, such as for example echo chambers, self-deception, deception symbiosis, and freezing of group opinions. As a result, the framework we propose could be used to develop methods to mitigate the impact of harmful communication strategies, i.e. in social media. We illustrate the potential for this via simulation experiments.