Automated agents' behavior in the Trust-Revenge game in comparison to other cultures

Agents that interact with humans are known to benefit from modeling them. Therefore, when designing agents intended for interaction with automated agents, it is crucial to model the other agents. However, little is known about how to model automated agents and in particular non-expert agents. Are automated agents to be modeled the same way that an agent models humans? Or does a separate model for interacting with automated agents need to be developed? We evaluate automated agent behavior (for non-expert agents) using a game called the Trust-Revenge game, which is known in social science for capturing different human tendencies. The Trust-Revenge game has a unique sub game-perfect equilibrium, however, very rarely do people follow it. We compared the behavior of automated agents to that of human actions in several demographic groups (including a group which is similar but not identical to the designers of the automated agents). We show that differences between automated agents' and humans' behavior are similar to differences between different human cultures.