Focus on What's Informative and Ignore What's not: Communication Strategies in a Referential Game

TL;DR

This paper investigates how neural agents develop communication strategies in a referential game, showing that environment distribution influences feature usage and language emergence.

Contribution

It reveals that non-uniform feature distributions lead to more efficient and natural language-like emergent communication strategies in neural agents.

Findings

Agents ignore less informative features in non-uniform environments.

Non-uniform environments improve the quality of emergent languages.

Agents rely on fewer features when some are less informative.

Abstract

Research in multi-agent cooperation has shown that artificial agents are able to learn to play a simple referential game while developing a shared lexicon. This lexicon is not easy to analyze, as it does not show many properties of a natural language. In a simple referential game with two neural network-based agents, we analyze the object-symbol mapping trying to understand what kind of strategy was used to develop the emergent language. We see that, when the environment is uniformly distributed, the agents rely on a random subset of features to describe the objects. When we modify the objects making one feature non-uniformly distributed,the agents realize it is less informative and start to ignore it, and, surprisingly, they make a better use of the remaining features. This interesting result suggests that more natural, less uniformly distributed environments might aid in spurring the emergence of better-behaved languages.