Constructing observational learning agents using self-organizing maps

Observational learning is a form of social learning whose theory proposes that new behaviors can be acquired through observing and imitating others. We employed Kohonen self-organizing maps to create observational learning agents to model the real-world process of observational learning. Real-world observational learning is a process that occurs through constantly changing nature and imperfect observation. In this study, we use observational learning agents to conduct a multiagent simulation of a cleanup problem comprising the chained tasks of picking up trash and subsequently discarding it. The results indicate that the constructed observational learning agents produce new emergent behaviors under changing, imperfect observation. Furthermore, the agents demonstrated the best performance when observing others to a moderate degree.