[3O1-OS-44a] OS-44

[3O1-OS-44a-04] Emergence of communication systems using multiple IBL agents in coordination task

〇Kenya Sasaki¹, Junya Morita¹ (1. Shizuoka University)

Keywords:Cognitive Model, Instance-Based Learning, Multi Agent, Communication

The importance of memory in the emergence of communication has been emphasized. Through repeated attempts at coordination, shared instances accumulate among members. These instances are abstracted through the human memory mechanism, forming a system that enables intentional communication. However, computational models based on this explanation vary depending on the adopted knowledge representation. This study compares two instance-based communication models: one using the integrated cognitive architecture ACT-R and the other based on PyIBL, which specializes in instance-based learning. Results show that communication formation is delayed in the PyIBL-based model compared to the ACT-R-based model with explicit goal representations and control structures. The delay is attributed to the difficulty in distinguishing between self and others’ memories in turn-taking situations in the PyIBL-based model. These findings suggest that the emergence of goal representations and the distinction between self and others may co-evolve with the process of language evolution.