Emergent communication in human-machine games

Nicolo' Brandizzi, Luca Iocchi

Keywords: Emergent Communication, Multi Agent Systems, Reinforcement Learning

Abstract:

In this paper, we present how Emergent Communication (EmeCom) literature can be used to characterize a subset of human communication ability using recurrent features.
EmeCom does not directly require human involvement; however, we advocate that the entire field spawned from a desire to achieve human-machine interaction and that explicitly considering human-AI interaction is an important research direction. Our goal is to bring emergent communication researchers together by demonstrating that different approaches can be regarded as subsets of the same problem. Also, we recognize the importance of incorporating other aspects of human interaction into this field.

TL;DR: We carry out an analysis of the current EmeCom literature where we promote the idea that this field has spawned by the necessity for human-robot collaboration; thus, the literature can be seen as tackling diverse aspects of human-human interactions.