Social norms learning in agent societies through reward or penalty observations have become the subject of interest in many studies. However, very few studies have examined the optimal environmental settings that would allow agents to learn through such observations effectively. This study presents a combination of environmental simulation parameters to discover the optimal settings for observing reward or penalty events, which are called the exceptional events, within a social agent group. The environmental settings consist of several variables which are the cycle time, observation limit of detector agent, domain size, population density of domain agents and occurrence of reward or penalty (exceptional) events in the domain. The value of each variable is arbitrarily set to low, medium or high. To implement the simulation, a virtual environment has been created with the variables settings to examine different situations. Within the steps of the tests, some cases are excluded because they do not significantly contribute to optimal environment for social learning. The results of the tests show that each variable has different effect on the environment and that a variable that has a strong positive effect does not individually offer the optimal solution. However, combining variables that have strong positive effects could offer optimal solutions. Briefly, the study aims to examine and identify the effect of some environmental variables on observation process of exceptional events and suggests the optimal settings to learn through observation.
All Science Journal Classification (ASJC) codes
- Artificial Intelligence